university of calgary assessing the effectiveness of wind...
TRANSCRIPT
UNIVERSITY OF CALGARY
Assessing the Effectiveness of Wind Power and Cogeneration for Carbon
Management of Electric Power Systems
by
Ganesh Doluweerawatta Gamage
A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF ELECTRICAL & COMPUTER ENGINEERING
CALGARY, ALBERTA
SEPTEMBER, 2011
© Ganesh Doluweerawatta Gamage 2011
Abstract
Climate change is an important environmental issue that may have significant
adverse impacts on human welfare. Consequently, prudent actions are needed
immediately to control the emissions of CO2, the main contributing greenhouse
gas for climate change. Since electric power generation is a location specific and
intensive source of CO2 emissions, focusing on reductions in this sector can have
relatively rapid and significant effects on overall CO2 levels. In this context, this
thesis makes three principal contributions.
First, the effectiveness of wind power for carbon management of electric power
systems is assessed. Wind power is considered as a critical technology to produce
electricity without CO2 emissions. However, wind power is a variable source of
electricity and power systems operated with wind power must ensure system relia
bility by firming wind variations. In this work, using the Alberta electric system as
a case study, the effectiveness of wind power for carbon management is assessed
taking the ancillary costs and emissions associated with firming wind power varia
tion into account. Operations of the Alberta electric system with different levels of
wind penetration are simulated using dispatch models that have sufficient resolu
tion to capture the dynamics of wind variations. The main result of this work is a
set of carbon abatement supply curves.
Second, a stochastic decision support model that can be used for operational
decision making of a wind power plant (WPP) and a compressed air energy storage
system (CAES) that jointly participate in a dayahead electricity market is devel
oped. This model inherently takes the uncertainty of wind and market price of
electricity and derives the optimal operating rules that maximize profits of the WPP
and CAES system.
ii
Third, the role of cogeneration for carbon management is evaluated utilizing
a mass and energy balance model and engineering economic analysis. By using
cogeneration for satisfying the energy demands of the oils sands operations in Al
berta as an example case, this work further examines the arbitrary characteristics
of facility and productbased carbon emissions control regulations.
Contributions of this thesis are intended to support efficient climate change
mitigation policy making.
iii
Acknowledgements
First, I convey my sincere thanks to my supervisors, Dr. Dave IrvineHalliday and
Dr. David Keith. It has been a privilege to carry out my doctoral research under
their supervision. I am thankful to Dave for all his support and advice, not only for
my graduate research work, but also for all of my endeavours at the University of
Calgary. David, with his deep knowledge in so many fields, helped me to shape up
both my research and teaching skills. Throughout my doctoral studies, David has
been truly an inspiration and I will always look up to him.
I am grateful to Dr. Michal Moore for his guidance, support and mentorship.
Michal has generously contributed an enormous amount of time and effort to en
sure my academic success.
I am indebted to Dr. Joule Bergerson and Dr. Bill Rosehart. I have benefit
ted tremendously from the guidance and support they rendered for my doctoral
research work.
I thank Dr. Janne Kettunen, Dr. Hamid Zareipour, and Dr. Ed Nowicki for
sharing their knowledge and making time for insightful discussions. I would like to
thank Ed also for his support in fulfilling my teaching responsibilities. His passion
for teaching has inspired me to become a better educator.
During my doctoral studies, I had the pleasure of working and interacting with
an exceptional group of students at the Energy and Environmental Systems Group
and the Department of Electrical and Computer Engineering. I specifically want to
thank my colleagues John MacCormack, Eduard Cubi, Mike Gestwick, Sarah Jor
daan, Nicolas Levy, Geoff Holmes, Hossein Safaei, Graeme Marshman, Mahmoud
Mazadi, and Amir Motamedi. The fruitful discussions I had and the collaborative
work I did with them supplemented my studies at the University of Calgary.
iv
I would like to thank the administrative staff of the Department of Electrical
and Computer Engineering and the Energy and Environmental Systems Group for
their assistance for my graduate studies. I particularly want to acknowledge the
support I received from Pauline Cummings, Ella Lok, Shannon Katusa, and Hollie
Roberts.
I acknowledge the partial financial support received for my doctoral research
through grants and scholarships from the University of Calgary, the Institute for
Sustainable Energy, Environment, and Economy, Natural Science and Engineering
Research Council, LCAOST Research Group, and the Canadian Association for
Energy Economics.
I acknowledge many friends and relatives who have encouraged and supported
my studies. I wish to specifically thank my friends, Arjuna Madanayake, Thushara
Gunaratne, Jithra Adikari, Gayan Wijesekara, and Lakshan Wasage for all the help
throughout my stay in Calgary.
Finally, last but not least, I am deeply grateful to my parents, my wife Yamuni,
my sister Anjana, and brotherinlaw Salinda. Their unconditional love, support,
and encouragements have been an important factor for the success in all aspects
of my life.
Ganesh Doluweerawatta Gamage
September 2011
Calgary AB, Canada
v
To my parents.
vi
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xNomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Climate Change and the Electric Power Sector . . . . . . . . . . . . . . . . . 21.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Assessing the Effectiveness of Wind Power for Carbon Management of Electric
Power Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.1 Introduction and Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Wind Power Integration and Carbon Management: Previous Studies . . 82.1.2 Research Objectives and Contributions . . . . . . . . . . . . . . . . . . 13
2.2 Description of the Simulation Experiment . . . . . . . . . . . . . . . . . . . . 142.2.1 Power System Studied in the Experiment . . . . . . . . . . . . . . . . . 16
2.3 Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.3.1 Unit Commitment Model . . . . . . . . . . . . . . . . . . . . . . . . . . 182.3.2 Realtime Operation Model . . . . . . . . . . . . . . . . . . . . . . . . . 212.3.3 Data and Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.3.4 Model Implementation and Simulation Workflow . . . . . . . . . . . . 27
2.4 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.4.1 Carbon Abatement Cost . . . . . . . . . . . . . . . . . . . . . . . . . . 302.4.2 Cost of Wind Variability . . . . . . . . . . . . . . . . . . . . . . . . . . 352.4.3 Transmission System Impacts . . . . . . . . . . . . . . . . . . . . . . . 442.4.4 Caveats and Limitations of the Study . . . . . . . . . . . . . . . . . . . 45
2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473 Risk Averse Shortterm Operations Optimization of Wind Power and Com
pressed Air Energy Storage Systems . . . . . . . . . . . . . . . . . . . . . . . 493.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.1 Dayahead Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.1.2 Compressed Air Energy Storage Systems . . . . . . . . . . . . . . . . . 523.1.3 Contributions of the Chapter . . . . . . . . . . . . . . . . . . . . . . . 53
3.2 Operations Optimization Under Uncertainty: Problem Description and Solution Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553.2.1 Stochastic Programming Solution . . . . . . . . . . . . . . . . . . . . . 573.2.2 Risk Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3 Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613.4 Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.4.1 Wind and Price Scenario Generation . . . . . . . . . . . . . . . . . . . 643.4.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.4.2.1 Risk Averse Operation . . . . . . . . . . . . . . . . . . . . . . . 693.4.2.2 Expected Value of Perfect Information . . . . . . . . . . . . . . 723.4.2.3 Sensitivity of CAES Parameters . . . . . . . . . . . . . . . . . 73
vii
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754 Evaluating the Role of Cogeneration for Carbon Management in Alberta . . . 774.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2.1 Oil sands operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.2.2 Alberta electric power system . . . . . . . . . . . . . . . . . . . . . . . 814.2.3 Current carbon management policies in Alberta . . . . . . . . . . . . . 83
4.3 Model Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.4 Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.4.1 CO2 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.4.2 Policy Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1034.4.3 Policy Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1105 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116A List of Power Generating Units . . . . . . . . . . . . . . . . . . . . . . . . . . 129B Optimal Operation of Standalone Wind Power Generation System . . . . . . 131C SGER Obligations Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . 133D Alberta Grid Average and Marginal Emissions Intensity Calculations . . . . . 136E CO2 Emissions Forecast of the Alberta Electric System . . . . . . . . . . . . . 139
viii
List of Tables
2.1 Generating units available at each bus (in MW) . . . . . . . . . . . . . . . . . 242.2 Generating unit heat rates and ramp rates . . . . . . . . . . . . . . . . . . . 242.3 Fuel prices and carbon intensities . . . . . . . . . . . . . . . . . . . . . . . . 272.4 New SCGT capacity required at different wind penetration levels . . . . . . . . 33
3.1 Model Parameters Used for the Numerical Example . . . . . . . . . . . . . . . 643.2 Expected profits, imbalance charges, and risk measures . . . . . . . . . . . . 673.3 Expected value of perfect information (EVPI) for the wind+CAES system . . . 72
4.1 Electricity and natural gas demand for bitumen extraction and upgrading. . . 804.2 Parameters used for the energy and CO2 emissions calculations. . . . . . . . 864.3 Cost parameters used for engineering economic analysis (all costs are in 2008
Canadian dollars). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.4 Cost of abated carbon emissions . . . . . . . . . . . . . . . . . . . . . . . . . 101
A.1 Power generating units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
C.1 Mass and energy balances of the ‘‘Baseline option" . . . . . . . . . . . . . . . 133C.2 Mass and energy balances of the ‘‘Cogeneration option" . . . . . . . . . . . . 133C.3 SGER obligations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
D.1 Alberta’s electricity production by generation technology (in GWh) . . . . . . 136D.2 CO2 intensity of the generation technology (in tCO2/MWh) . . . . . . . . . . . 136D.3 Percentage of the time different generation technologies set the price in Al
berta’s whole sale electricity market. . . . . . . . . . . . . . . . . . . . . . . . 138
E.1 Installed electricity generation capacity in Alberta (in MW) . . . . . . . . . . . 140E.2 Generation Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
ix
List of Figures
2.1 Simplified transmission model of the Alberta Electric Power System . . . . . . 172.2 Demand duration curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3 Electricity produced by different generation technologies. . . . . . . . . . . . 292.4 Average CO2 intensity of the energy mix. . . . . . . . . . . . . . . . . . . . . . 302.5 Carbon emissions intensity and average cost of electricity . . . . . . . . . . . 322.6 CO2 abatement supply curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.7 Cost of wind uncertainty and variability . . . . . . . . . . . . . . . . . . . . . 362.8 CO2 emissions stem from mitigating uncertainty and variability of wind power. 392.9 Correlation between intertime step ramps of WPPs and other dispatchable
generating units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402.10Coal unit operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422.11Power flow duration curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.1 CAES system configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533.2 Systems under study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563.3 Operations decisions time line . . . . . . . . . . . . . . . . . . . . . . . . . . 563.4 Two stage decision making process . . . . . . . . . . . . . . . . . . . . . . . . 583.5 βVaR and βCVaR of a profit distribution Bs . . . . . . . . . . . . . . . . . . 603.6 System price and wind power scenarios used for the case study . . . . . . . . 663.7 Profit distributions of: (a) standalone wind power plant operation; (b) Wind +
CAES joint operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683.8 Energy bids to the dayahead electricity markets . . . . . . . . . . . . . . . . 693.9 Operation of the CAES system under the scenario #48. . . . . . . . . . . . . 703.10Efficient frontiers of the two power plant configurations . . . . . . . . . . . . 713.11Influence of CAES parameters on Wind+CAES system economics . . . . . . . 74
4.1 (a) Baseline option and (b) cogeneration option . . . . . . . . . . . . . . . . . 854.2 Total CO2 emissions within Alberta, under the two energy options . . . . . . 934.3 CO2 emissions intensities of electricity . . . . . . . . . . . . . . . . . . . . . . 954.4 Forecast of CO2 emissions from the Alberta electric system to 2020 . . . . . . 984.5 Forecast #2 of CO2 emissions from the Alberta electric system to 2020 . . . . 994.6 Emissions reductions obligations . . . . . . . . . . . . . . . . . . . . . . . . . 1044.7 Emissions offset credits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
D.1 Average CO2 intensity of the Alberta Grid in years 2000 to 2008 . . . . . . . . 137D.2 Marginal CO2 intensity of the Alberta grid . . . . . . . . . . . . . . . . . . . . 138
E.1 Forecasted installed generation capacity in Alberta (20092020) . . . . . . . . 141
x
xi
Nomenclature
Abbreviations
AESO Alberta Electric System Operator
BEI Baseline emissions intensity
CAD Canadian dollars
CAES Compressed air energy storage
CCEMF Climate Change and Emissions Management Fund
CCGT Combined cycle gas turbine
CCS Carbon capture and sequestration
CO2 Carbon dioxide
CO2 eq. Carbon dioxide equivalent
CSS Cyclic steam stimulation
CVaR Conditional value at risk
DAM Dayahead market
EES Electric energy storage
EPC Emissions performance credits
EVPI Expected value of perfect information
EWS expected value of wait and see solution
FCE Fuel chargeable to electricity
GHG Greenhouse Gas
HHV Higher heating value
HRSG Heat recovery steam generator
IEA International Energy Agency
IPCC Intergovernmental Panel for Climate Change
ISO Independent System Operator
LCA Life cycle assessment
MAE Mean absolute error
MCR Maximum continuous rating
xii
PHS Pumped hydro storage
RO Realtime operation
RTO Regional Transmission Operator
SAGD Steam assisted gravity drainage
SCGT Simple cycle gas turbine
SCPC Supercritical pulverized coal
SGER Specified gas emitters regulation
SP Stochastic programming
TSO Transmission system operator
UC Unit commitment
VaR Value at risk
WPP Wind power plant
Sets and Indices
R Set of real numbers
G, j Set and index of generating units
Gk Set of generating units connected to bus k
Gh Set of hydro generating units
Gmr Set of mustrun generating units
Gth Set of thermal generating units
Gw Set of wind generating units
K, k Set and index of buses
S, s Set and index of scenarios
T, t Set and index of time
Parameters
ηB Baseline boiler efficiency
ηc CAES electricity input/output ratio
ηG Heat recovery steam generator (HRSG) supplemental firing efficiency
ηR HRSG heat recovery efficiency
ηT Electricity generation efficiency of the gas turbine
xiii
λ Penalty factor for energy imbalance
πst Market price under scenario s in hour t [$/MWh]
ρs Probability of scenario s
τ Optimization time step (=1h)
τ rt Simulation time step of the realtime operation model (=10mins)
τuc Simulation time step of the unit commitment model (=1h)
τx Exergetic temperature factor under assumed steam conditions
ξ−i Binary parameter that represent the offline history of unit j
ζ−i Binary parameter that represent the running history of unit j
bkr Susceptance of the transmission line between buses k and r
cf CAES fuel cost [$/MWh]
C ls Cost of unserved demand [$/MWh]
Csuj Start up fuel cost of unit j [$]
Cmin, Cmin Minimum and maximum limit of the CAES system’s air compressor [MW]
Ddnj Minimum down time of unit j [h]
Dupj Minimum up time of unit j [h]
E Onsite electricity demand (MWh/h)
Emin, Emin Minimum and maximum limits of the air storage cavern [MWh]
H Onsite steam demand / steam produced by baseline boiler (GJ/h)
Icng CO2 intensity of natural gas (tCO2/GJ)
Icogen CO2 intensity of cogenerated electricity (tCO2/GJ)
lkt Power demand at bus k in time period t [MW]
P havit Hydro resource availability for the unit i in time period t [MW]
Pminj , Pmax
j Minimum and maximum power generation capacity of unit j [MW]
Pmrjt Mustrun capacity of the unit j in time period t [MW]
P resj Maximum reserve power provided by unit j [MW]
P rlj Ramping limit of the unit j [MW]
Pwavit Wind resource availability for the unit i in time period t [MW]
pc Price of carbon ($/tCO2)
xiv
pe Price of electricity ($/MWh)
ph Value of steam ($/GJ)
Pmin, Pmin Minimum and maximum power generation limits of the CAES system [MW]
Pramp Maximum allowable ramp rate [MW/h]
Ptx Maximum available transmission capacity [MW]
Pwmax Maximum power generation limit of the wind farm [MW]
Rt Reserve power requirement of the system in time period t [MW]
T Temperature of steam (K)
Tmaxkr Transmission capacity of the line between buses k and r [MW]
T0 Temperature of the reference environment (K)
Wst Wind power generation under scenario s in hour t [MWh]
u System utilization (assumed to be 90%)
Variables
αj , αw Profit threshold levels of joint operation and sandalone wind operation [$]
δkt Voltage angle of bus k in time period t [rad]
bjt Joint energy bid to the day ahead market by the wind farm and CAES system
in hour h [MWh]
bwt Energy bid to the day ahead market by the wind farm in hour h [MWh]
Bjs Profit function of wind+CAES joint operation [$]
Bws Profit function of standalone wind operation [$]
dst Compression power of the CAES system in hour h under scenario s [MW]
EC Electricity produced by the cogeneration system (MWh/h)
Eexp Electricity exported to the grid (MWh/h)
FB Fuel input to the baseline boiler (GJ/h)
FG Fuel input to the HRSG of the cogeneration system (GJ/h)
FT Fuel input to the gas turbine of the cogeneration system (GJ/h)
FSB Fuel input to the supplementary boiler (GJ/h)
FCE Fuel chargeable to electricity (GJ/MWh)
gcst Power output of the CAES system in hour h under scenario s [MW]
gwst Power output of the wind farm in hour h under scenario s [MW]
gjt Power output of the unit j in time period t [MW]
H1 Steam produced by cogeneration system (GJ/h)
H2 Steam produced by auxiliary boiler (GJ/h)
Hfw1,Hfw2 Cogeneration system / auxiliary boiler feed water enthalpy (GJ/h)
Hfw Baseline boiler feed water enthalpy (GJ/h)
llskt Unserved power demand at bus k in time period t [MW]
mst, nst Binary variables that represent the operating mode of the CAES system inhour h under scenario s
pkt Summation of power flow in all transmissions lines between bus k and allother buses in time period t [MW]
rjt Reserve power capacity provided by unit j in time period t
rst Energy stored in the storage cavern in hour h under scenario s [kWh]
ujt Commitment of unit j in time period t (binary variable)
yjt Shut down of unit j in time period t (binary variable)
zjt Start up of unit j in time period t (binary variable)
xv
1
Chapter 1
Introduction
Climate change is one of the most important environmental issues that humanity
confronts. Scientific studies such as the assessment reports of the Intergovernmen
tal Panel for Climate Change (IPCC) [1] conclude that climate change is unequivocal
and driven mainly by human activities. A primary concern is the climate anomalies
triggered by rising average atmospheric temperature due to the increased levels of
greenhouse gases (GHG) in the atmosphere [2]. While there are many GHGs, an
thropogenic carbon dioxide (CO2) is the main contributor for the climate change.
The main source of anthropogenic CO2 is fossil fuel combustion for energy
services; with this the post industrialized era saw a steady increase in CO2 con
centration [2]. For example, fossil fuel CO2 emissions has increased from 25.7
MtCO2/year in 2000 to 30.8 MtCO2/year in 2006. The average growth rate of fossil
fuel CO2 emissions has increased from 1.3%/year for 19901999 to 3.3%/year for
20002006 [3]. Scientific studies based on the present understanding of the earth’s
climate system have predicted a range of adverse impacts of climate change, includ
ing events with direct impacts on humanity such as precipitation pattern changes,
decrease of agricultural productivity, heat waves, and catastrophic events such as
sea level rise due to the complete melting of west Antarctic ice sheets [2,4]. Due to
the complex nature of the earth’s climate system, there is considerable uncertainty
in these impacts [1, 2, 5]. Nevertheless, the potential magnitude of the impacts of
such events on human welfare and the long timescales that involve in the develop
ment of new energy infrastructure and in the climate response for changes in GHG
emissions call for immediate implementations of climate change mitigation strate
2
gies, mainly those that are directed to reduce anthropogenic GHG emissions [2,5].
For example, Caldeira et al. show that regardless of the range of uncertainty for
climate sensitivity1, as predicted by present climate models, 75 to 100% of human
ity’s total energy demand will need to be provided by non CO2 emitting sources by
the end of twentyfirst century in order to stabilize the earth’s average temperature
at 2◦C above the pre industrial levels2 [7]. A recent study by Meinshausen et al. [6]
has developed probability distributions for exceeding earth’s average temperature
by 2◦C above the pre industrial levels under various CO2 emissions scenarios in
the first half of the twentyfirst century (20002049). According to that study, in
order to limit the probability of exceeding 2◦C warming below 25%, the cumula
tive CO2 emissions in the 20002049 period should be kept below 1000 GtCO2 [6].
However, as discussed in [3] the cumulative emissions in 20002006 period were
234GtCO2 with an emissions rate of 36.3GtCO2/year. According to [6], keeping
the emissions rate at the 2006 level for the remainder of the century will make the
probability of exceeding 2◦C warming greater than 50% by 2039. As a consequence,
prudent actions must be taken immediately to control the CO2 emissions to reduce
the likelihood of adverse impacts on human wellbeing due to climate change.
1.1 Climate Change and the Electric Power Sector
The electric power sector is the largest CO2 emitter of the world due to its fossil
fuel dominated electricity generation. In 2008, approximately 68% of the primary
energy used by the electricity sectors worldwide was fossil fuels, of which 41 per
centage points was high CO2 intensive coal [8]. The high share of emissions and
the fact that electric power plants are large point CO2 sources have made the elec
1Climate sensitivity is defined as the global mean climatological temperature change resultingfrom a doubling of atmospheric CO2 content
2Warming limit of 2◦C or bellow has been set by over 100 countries as a guiding principal formitigation efforts [6]
3
tric power sector the main target for CO2 emission reduction. Furthermore, the
marginal cost of reducing emissions in the electricity sector is reported to be lower
than other sectors such as transportation, and therefore, the electricity sector
may deliver the largest proportional CO2 reductions under an economically efficient
climate policy [9]. For example, emission reduction scenarios developed by the
International Energy Agency (IEA) to stabilize the atmospheric CO2 concentration
at 450ppm by 2050 calls for more than 50% CO2 reduction in the electric power
sector [10].
An ensemble of proven technology options for managing CO2 emissions from
electric power sector exist; most of them are already in use at different scales.
These options include nuclear power, hydroelectric power, biomass, geothermal,
wind power, solar power, and fossil based generation with carbon capture and se
questration (CCS) [9–12]. The renewable power options such as hydro, solar, and
wind are considered the most environmentally benign options for producing elec
tricity without carbon emissions and they are receiving increased interest world
wide. At present approximately 19% of primary energy consumed by the electricity
sectors worldwide are renewables, of which approximately 85% is hydropower [8].
The principal objective of this thesis is to assess technical and policy options
for carbon management of electric power systems.
1.2 Research Objectives
Decarbonizing the electricity sector requires investments in low/zero carbon in
tensive electricity generation technologies. Those investments are complicated by
the existing capital stock of power generation infrastructure, which is long lived
and implies significant replacement costs. Therefore, effective carbon management
policies are required to attract investments in low carbon intensive generation tech
4
nologies. To achieve optimal results, carbon management policies should be de
signed such that, when enforced, the generation options that provide maximum
CO2 emissions reduction at the lowest cost becomes economically competitive, at
tracting investment. Among other things, a key information requirement for the
design of such policies is the marginal abatement costs of carbon of different op
tions available for mitigating CO2 emissions.
The first objective of this thesis is to provide high confidence estimate of the
carbon abatement cost of wind power. Wind power is considered to be a critical
technology to move away from fossil fuel derived electricity and to sustain deep
carbon cuts in the electricity sector.
Estimating the abatement cost of wind power is complicated by the variable
nature of wind power generation. In power system operations, the demand and
supply need to be matched almost instantaneously. Therefore, in order to operate
electric power systems with a large amount of wind power, firming power has to be
provided from dispatchable generating units, which may add costs and emissions.
In this thesis, the carbon abatement cost and variability cost of large scale wind
power are estimated through simulations of real power system operations.
The second objective of this work is to investigate the feasibility of using large
scale electric energy storage (EES) systems to mitigate the challenges faced by
a wind power plant (WPP) operator due to variability of wind power in practical
power system environments. Compressed air energy storage (CAES) is a proven
large scale EES technology that has gained interest of power system planners and
investors. In this thesis, an operations optimization model that can be used to
support optimal operations decisions of a WPP that is in joint operation with a
CAES system is developed. The model specifically addresses the uncertainty in the
wind resource availability and market price of electricity.
5
The third objective of the thesis is to provide insights into the arbitrary nature
of some present carbon management policy practices. An economically efficient
climate policy should create a single economy wide marginal carbon price signal
either in direct form, such as a carbon tax, or in an implied form such as a cap and
trade system. However, in response to political pressure against enforcements of
such measures, some jurisdictions have chosen to use complex facility or product
based policy tools to control carbon emissions. However, the choice of facility or
productbased carbon accounting methods is inherently arbitrary. In this thesis,
analysis of emissions accounting methods and insights into their arbitrariness
are provided by taking an illustrative example from the use of cogeneration for
electricity and heat production for oil sands operations in the Canadian province
of Alberta.
The research work presented in this thesis is designed to contribute to the
engineering and public policy knowledge pertaining to carbon management policy
making.
1.3 Thesis Structure
The rest of the thesis is organized as follows.
Chapter 2: The research work presented in this chapter assesses the role of large
scale wind power for carbon management of electric power systems. Carbon
abatement supply curves of wind power pertaining to the Alberta electric
system are developed by simulating the power system operations in a single
year at 10 minutes time steps. This chapter provides analysis designed to
support efficient carbon management policy. A review of related literature
and mathematical formulation details of the models used to simulate the
Alberta electric system are also presented.
6
Chapter 3: This chapter presents an operations optimization model for a wind
power plant and a CAES system that jointly participate in a dayahead elec
tricity market. The model utilizes a two stage stochastic programming ap
proach to produce optimal operations decisions under wind resource and
electricity price uncertainty. The model also includes integrated risk man
agement measures.
Chapter 4: This chapter evaluates the role of cogeneration as a carbon manage
ment option for Alberta and provides insights about the arbitrariness of the
choice of accounting methods for facility based carbon management policies.
This chapter also evaluates the efficiency of current and alternative emissions
control policies in Alberta.
Chapter 5: This chapter summarizes the conclusions and contributions of the
thesis.
7
Chapter 2
Assessing the Effectiveness of Wind Power for Carbon
Management of Electric Power Systems
2.1 Introduction and Background
Wind power is currently the fastest growing electricity generation technology in
the world. Since 2001, global installed wind capacity has grown by 2040% per
year. At the end of 2010 the global installed wind capacity was 197GW, which is
approximately 11 times the capacity in 2000. Five countries namely, Germany,
United States, Spain, India, and China hold 73% of global installed wind capacity
[13]. U.S. had the highest installed capacity at 40.2GW and China had the highest
growth rate. As of April 2011, Canada’s installed wind capacity was 4588MW with
Ontario leading at 1636MW, followed by Alberta at 777MW [14]. The average wind
power growth rate in Canada in 20062010 was 29%.
Since the early 1980s commercial wind turbines have improved enormously
in terms of their capacity, efficiency, and visual design. The cost of wind power
has been reduced by a factor of four since the 1980s, largely due to the technol
ogy developments, scaling up turbine size, and increased manufacturing capac
ity [10, 15,16]. However, the turbine prices have increased since 2004, with high
demand being a driving factor [10, 16]. The cost of wind generated electricity is
largely determined by available wind resources. According to estimates presented
in [10], accounting for the capital costs and operating and maintenance costs, the
cost of electricity generated from wind is 89135 US$/MWh for low average wind
speeds and 6094 US$/MWh for high average wind speeds. Many jurisdictions pro
8
vide incentives and production credits as well as supporting policy environments
for wind power developments. Some notable examples include the U.S. Federal Pro
duction Credit, Renewable Portfolio Standards adopted by various U.S. states, fixed
feed in tariff system in Germany, European Union Renewable Electricity Directive,
and EcoEnergy clean energy production credits in Canada [16–19].
Wind power has emerged as a proven technology for obtaining deep carbon cuts
in the electric power sector as well as to ensure energy security by an ensemble
of future generation scenarios [10, 11,20–22]. However, two factors—the variabil
ity of wind power generation and the spatial distribution of wind resources—may
complicate the effectiveness of wind power to reduce carbon emissions in the elec
tric power sector [23]. The research study presented in this chapter assesses the
effectiveness of large scale wind power for carbon management of electric power sys
tems. The remainder of the chapter is organized as follows. Section 2.1.1 describes
the research problem and reviews the related literature. Section 2.1.2 formulates
the research questions. Sections 2.2 and 2.3 describes the simulation experiment
carried out and the mathematical formulation of simulation models. Results of the
study are discussed in section 2.2 and the conclusions are drawn in section 2.5.
2.1.1 Wind Power Integration and Carbon Management: Previous Studies
To ensure the reliability of an electric power system, electricity demand and supply
must be matched instantaneously as large scale electric energy storage is presently
not widespread available. Since demand is variable, in the process of matching de
mand with supply, variability in three time scales must be addressed by the power
system operator: secondstominutes, intrahour, and hours to days [24]. Hours
to days demand variations are managed using demand forecasts, unit commitment
services, and economic dispatch services. Variations in the remaining two time
scales are managed using ancillary services. Regulation service that employs au
9
tomatic generation control is employed to manage secondstominutes variations.
Load following service that use operating reserves (both spinning and nonspinning)
is used to manage intrahour variations1.
Wind power is a variable source of power and it is difficult to predict the wind
resource availability with high accuracy over daily periods [25]. The output of
a wind power plant (WPP) typically varies over time scales of seconds to weeks
[26]. Hence, adding WPPs significantly increase the variability of the supply side
in all three aforementioned time scales. In order to compensate for wind power
variations, system operators must procure firming capacity through regulating and
operating reserves, increasing the operating costs. Furthermore, the operating
costs can increase due to impacts on other generating units. For example, to be able
to ramp up or down to firm up the wind variations, thermal generating units in the
system may have to divert from their optimal and most efficient operating points,
increasing fuel costs. Number of conventional unit startups and shutdowns may
increase (both planned and unplanned), in order to maintain committed capacity to
manage system variability. Since thermal generating units in general are designed
for continuous operations the cyclical operations due to wind variability may lead
to higher wear and tear, increasing the maintenance costs [27]. Other cost factors
include high operating costs of fast response units that are dispatched to firm
up wind power variations, lost revenues for conventional generating units, and
demand curtailments.
The impacts of wind power variability on power system operations have been
previously studies. The vast majority of these studies have been carried out by
Independent System Operators (ISO) and Regional Transmission System Operators
1The exact terminology of these ancillary services and the time boundaries that separate themare not universal; depend on the power system that employs them. Discussion presented here hastaken the most common terminology in power system literature
10
(RTO) of power systems with increasing wind penetration levels2. A number of
wind integration studies from the U.S. are summarized in [16,24]. Similar studies
from some European countries are summarized in [28,29]. The impacts of wind
power on the Ontario power system are studied in [30] and on the Alberta power
system in [31–33]. All studies cited here have assessed the physical impacts of
wind power on power system operations; some have also estimated the financial
impacts. Methods employed for these studies include probabilistic methods such
as ones proposed in [34] and sophisticated dispatch models and electricity market
simulators, that are usually proprietary and accessible only by ISOs and RTOs.
The physical impacts of wind, as found by the aforementioned studies, include
increase of regulating and load following reserve requirements and need for flexi
ble generating units with higher ramp rates. The costs of wind power variability
reported in the U.S. studies are in the range of 0.459 US$/MWh of wind energy
at wind penetration levels of 3.531%. The exact variability cost depends on the
study assumptions and the structure of the power system being studied. Neverthe
less, the general conclusion is that the variability cost rises with increasing wind
penetration level.
Another issue with wind power is that the sites with good wind resources gen
erally tend to be in remote locations, far from the major demand centers. This fact
has been observed in almost all jurisdictions worldwide that experienced signifi
cant wind power developments [23,35]. Grid integration of wind power plants sited
on those sites demands new transmission development and/or reinforcement of
existing transmission systems, increasing the cost of wind integration [36]. WPPs
produce electricity only when wind is blowing and the reported capacity factors for
onshore wind power plants are in the range of 2535% [10,37]. This characteristic
2Throughout this thesis the term wind penetration refers to the installed wind capacity as apercentage of the annual peak demand of the power system
11
may lead to transmission line underutilization and higher transmission cost.
The variability of wind resources also leads to uncertainties in the effectiveness
of wind power for carbon management. Regulation and load following reserves
that are employed to provide firming power are, in general, conventional thermal
generating units with fast response capability, although hydroelectric generators
are the preferred choice where available. The main types of conventional generators
used are natural gas fired simple cycle and combined cycle gas turbine generating
units (SCGT and CCGT), resulting in nonzero net CO2 emission from wind power.
Moreover, in order to firm the output of a WPP, the firming units may have to be
ramped up/down more frequently. As a consequence a recent study that simulated
WPPs backed up by a SCGT units in intrahour time scale, concludes that the SCGT
units may emit higher than expected levels of CO2 emissions due to the anomalous
ramping [38]. The same study also found higher levels of nitrogen oxides (NOx), a
regulated class of air pollutants, than their rated emission levels due to high unit
ramping.
Electricity generation units displaced by wind power have a significant impact
on CO2 emissions reduction in electric power systems. For example, wind power,
particularly at low penetration levels, may displace only marginal generating units.
To satisfy the time varying demand, power system operators typically dispatch dif
ferent generating units by taking a cost minimization approach. Usually marginal
units are natural gas fired generators that have higher marginal costs but lower
carbon intensities. High carbon intensive coal units usually operate as baseload
units due to their lower marginal cost and may not be displaced by wind power.
However, this fact can change with higher wind power penetration levels.
Two other factors may lead to uncertainties in the effectiveness of wind power
for carbon management of electric power systems. First, in addition to increasing
12
the cost, operating thermal units at nonoptimal points can increase the carbon
intensity of produced electricity due to reduced relative efficiency. Second, increase
in unit startups and shutdowns consume more fuel, increasing CO2 emissions.
The carbon management benefits of wind power have also been previously stud
ied. CO2 emissions reduction potential of wind power in Nordic countries is studied
using a commercially available electricity market simulators in [39]. Through power
system simulations, this study has estimated potential CO2 emissions reductions
in a single year (at weekly time steps) and multiyear periods (at 5year time steps).
The study has found that the emissions reduction potential of wind range from
0.620.7 tCO2/MWh of wind at an abatement cost of 3520 Euro/tCO2 (43.7525
US$/tCO2).
Carbon abatement supply curves3 for wind power in the U.S. and OECD Europe4
are developed in [40]. Aggregated load duration curves are used with a power
system simulator that runs in monthly time steps for this study. Carbon abatement
cost of 20155 US$/tCO2 has been reported to abate 50550 MtCO2.
CO2 , NOx, and SO2 (sulfur dioxide) emissions reduction potential as well as net
economic benefits of wind power in Ireland are assessed in [27] using a dispatch
model that runs at hourly time steps. One important finding of this study is the
potential increase in CO2 emissions at high wind penetration levels due to cyclical
operation of thermal units.
The economics of wind power considering both variability and spatial distribu
tion of wind resources have been studied in [23]. This study utilized a greenfield
approach where WPPs, CCGT units, SCGT units, compressed air energy storage
systems, and transmission lines are optimally built to satisfy an electricity demand.
3A carbon abatement supply curve plots the marginal abatement cost against the level ofCO2 emissions reduction achieved by an emissions mitigation option
4OECD Europe includes the Western European countries that participate in the Organisation ofEconomic Cooperation and Development
13
Using a dispatch model that runs in hourly time steps, this paper investigated the
CO2 emissions reduction potential of wind power. The study concludes that even
with the increase in average cost due to variability and transmission line addi
tions, wind power is a competitive option for managing carbon emissions in electric
power systems. This study also formulated supply curves of carbon abatements
and finds that 50% of emissions reductions relative to the baseline (the baseline
system consists of CCGT and SCGT units) can be achieved at an abatement cost of
125 US$/tCO2.
2.1.2 Research Objectives and Contributions
Reducing CO2 emissions in the electric power sector is one of the most significant
climate change mitigation options available to policy makers. Wind power is a
proven technology that can produce electricity without carbon emissions. As De
Carolis and Keith argue in [23], perhaps the most important role of wind power
is to produce electricity without CO2 emissions. However, as discussed in sec
tion 2.1.1, due to the spatial dispersion of wind resources, ancillary costs of wind
variability, and carbon emissions associated with managing the variability of wind
power, there is uncertainty in wind power’s ability to reduce CO2 emissions at a
competitive cost. It is imperative that the carbon abatement cost of wind power is
estimated at high confidence levels taking into account the challenges for practical
power systems due to the natural behavior of wind. Such estimates can be used
to formulate carbon abatement supply curves to inform the carbon management
policy making.
The principal research question posed in this chapter is: if a large volumes of
wind power were added to an existing electric power system, how would the carbon
abatement cost unfold ? The research work presented in this chapter attempts to
answer this question by simulating operations of the electric power system of the
14
Canadian province of Alberta.
Secondary objectives of the study are a) to estimate the cost of uncertainty and
variability of wind power at different penetration levels; b) to provide insights of the
transmissions system requirements necessary to support largescale wind power
developments in Alberta.
In contrast with previous studies, the key contribution of this work is formula
tion of carbon abatement supply curves that are generated using a high resolution
dispatch model, more effectively capturing the wind power and electricity demand
dynamics. System impact studies presented in [16,24,28–33] have focused only on
operational costs and do not provide emissions estimates. The penetration levels
considered for these studies are also limited (up to 2030%). Simulation models
in [39,40] relied on very long time steps and aggregated demand data. Therefore,
they were not able to capture the variability of wind power. In this respect, mod
els used in [23, 27] have made improvements by using an hourlong time steps,
although they may still not be able to fully capture the impacts of wind variability.
The simulation models developed for this study have very detailed represen
tations of generating units, including the efficiency degradations due to partload
operations of thermal generating units, startup fuel consumptions, ramping limits,
and challenges for operations planning due to unpredictability of wind resources
in daily time periods. Furthermore, the models and the data used for the current
work are fully transparent, which is important for analyses that are intended to
inform policy.
2.2 Description of the Simulation Experiment
In this section, a simulation experiment is designed to estimate the carbon abate
ment cost of largescale wind power. In order to make an accurate estimate of
15
operating costs and emissions, it is important that the simulation models cap
tures the dynamics of wind power and demand variations. It is also important
that the methodology used approximates how the actual power systems are op
erated. As discussed in section 2.1.1, the variations of demand and wind can
be divided in to three time frames: secondstominutes, intrahour, and hours to
days. This simulation experiment focuses on the latter two time frames. Impacts
in secondstominutes time frame are not studied, primarily due to unavailability
of highresolution wind power and demand time series data.
In power system operations, satisfying the demand at a certain point in time
is a multistage process. The system operators forecast the demand and schedule
generating units to satisfy the demand of all time periods in the planning period
at the lowest cost while taking physical constraints of the generating units and the
transmissions system into account. Prior scheduling of generating units must be
carried out for number of reasons, including the lead times of thermal generating
units to synchronize to the network and deliver power, minimum up time and
down time limits of thermal generating units, spinning reserve allocations to ensure
system reliability, and transmission system constraints. In case of deregulated
electric power systems, prior scheduling of generating units is done by conducting
an auction, where energy suppliers and buyers bid hourly prices and quantities (see
section 3.1.1). In realtime operation, the generating units are dispatched to satisfy
the demand at a lowest cost, with subject to physical limitations. The available
generating units to satisfy the realtime demand are limited by the schedule set in
the planning stage, and system operators maintain capacity reserves to respond
for unforeseen events. Unscheduled fast acting units may have to be brought
online. Operators of power systems with higher wind penetrations, may have to
supplement operations planning by forecasting the wind power availability.
16
In this experiment, two models–a unit commitment model and a realtime op
eration model–have been developed to simulate the supply and demand matching
process of a practical power system. The objective of the unit commitment model
(hence forth referred to as the UC model) is to schedule the available generating
units to satisfy the forecasted demand in the all time periods in the planning period
at a minimum operating cost. Further details of unit commitment problem can be
found in [41]. The UC model schedules the generation units for a period of 24 hours
to satisfy the forecasted load at 1h time steps. UC model also takes the forecasted
wind power availability in the planning horizon into account. The mathematical
formulation of the UC model is discussed in section 2.3.1.
The realtime operation model (henceforth referred to as the RO model) is formu
lated as an economic load dispatch (ELD) problem [41]. This model simulates the
real time generating unit dispatch in time steps of 10 minutes. It should be noted
that the RO model dispatches the generating units according to their marginal
cost. This is analogous to a fully competitive electricity market setting where none
of the generating units have market power. The mathematical formulation of the
RO model is discussed in section 2.3.2.
A similar multistage approach has been taken by Sioshansi & Short in [42]
to study the effect of realtime pricing on economics of wind power; by Wang et
al. in [43] to study the impact of wind power on thermal unit commitment and
dispatch.
2.2.1 Power System Studied in the Experiment
Due to the province’s need to reduce the CO2 emissions, the growing wind power
industry, and high share of coalfired generator fleet, this research study uses the
Alberta electric power system as the test system to investigate the effectiveness
of wind power for carbon management. The Alberta power system’s current total
17
(6) Northwest (5) Northeast
(4) Edmonton
(3) Central
(2) Calgary
(1) South
Figure 2.1: Simplified transmission model of the Alberta Electric Power SystemEach bus represents a transmission region in Alberta. All of the WPPs considered for thesimulation experiment are in bus 1 (South). The seven transmission lines correspond tomajor transmission corridors. Major demand centers are in regions represented by buses2 (Calgary), 4 (Edmonton), and 5 (Northeast).
installed generation capacity is 13,520 MW, of which approximately 46% is coal
fired generating units [44]. The system has a winter peak demand of approximately
10,000 MW and the annual load factor is about 80%. Alberta’s electricity trans
mission system consists of a 240kV backbone connected to 144kV, 138kV, 72kV,
and 69kV low voltage transmission lines and over 2000 buses [44]. The electricity
sector in Alberta is deregulated and power generation is treated as a competitive
business. The Alberta Electric System Operator (AESO) is the independent system
operator in Alberta. The current installed wind power capacity is 777MW and ac
cording to the AESO another 1600MW of WPPs have either applied to or received
power plant approval. The interest for potential wind power developments is re
ported to be about 6GW. Among other reasons, limited interconnections and high
share of inflexible generator fleet have posed challenges for large scale wind power
in Alberta. More than 60% of the annual electricity produced is generated by coal
18
fired power plants making the Alberta power system the most carbon intensive
electric power system in Canada. In 2008, electricity generation in Alberta pro
duced 53.2 MtCO2 equivalent, which is 7.3% of total Canadian emissions and 22%
of the provincial emissions [45].
The models developed for this study represent the Alberta power system by
6 buses and 7 transmission lines. Each bus represents a transmission region
as per AESO specifications [46]. This simplified transmission model is depicted
in figure 2.1. Alberta currently has limited tie line capacities to the neighbouring
power systems of British Columbia and Saskatchewan. These lines are not modeled
in the current implementation of the model. Exclusion of interconnections does not
significantly affect the results and conclusions of this study as imports and exports
represent only about 3% and 0.7% of the total demand of Alberta respectively. The
results produced by the model are crosschecked with actual historical data and
found to have satisfactory accuracy levels.
2.3 Model Formulation
2.3.1 Unit Commitment Model
The UC model is characterized by the objective function (2.1).
minimize∑
t∈T
∑
j∈G
[
fj(gjt) · τuc + Csu
j · zjt]
+∑
t∈T
∑
k∈K
C ls · luskt · τuc (2.1)
The objective function (2.1) minimizes the sum of fuel costs, unit startup costs
and unserved demand costs in planning horizon T. Cost summation is appropri
ately carried out over the set of generating units G, set of buses K and set of time
T. The fuel cost in supplying gjt MW by unit j in time period t is calculated by the
unit’s fuel cost function, fj(·) (in $/h). The time step of the UC model, τuc, is equal
19
to 1 hour. The startup cost of unit j is Csuj ($) and startup of unit j in time period
t is determined by the binary variable zjt. C ls is the cost of unserved demand (in
$/MWh) and luskt is the unserved demand (in MW) at bus k in time period t. The
operating cost is minimized subject to constraints (2.2)(2.16).
∑
i∈Gk
git = lkt − luskt + pkt, ∀k ∈ K, ∀t ∈ T (2.2)
pkt =∑
r∈K
−bkr · (δkt − δrt), ∀k ∈ K, ∀t ∈ T (2.3)
− Tmaxkr 6 −bkr · (δkt − δrt) 6 Tmax
kr , ∀(k, r) ∈ K, ∀t ∈ T (2.4)
ujt · Pminj 6 gjt + rjt 6 ujt · P
maxj , ∀j ∈ G, ∀t ∈ T (2.5)
∑
j∈G
rjt > Rt, ∀t ∈ T (2.6)
rjt 6 P resj , ∀j ∈ G, ∀t ∈ T (2.7)
git > Pmrit , ∀i ∈ Gmr, ∀t ∈ T (2.8)
git 6 P havit , ∀i ∈ Gh, ∀t ∈ T (2.9)
git 6 Pwavit , ∀i ∈ Gw, ∀t ∈ T (2.10)
zjt > ujt − ujt−1, ∀j ∈ G, ∀t ∈ T (2.11)
yjt > ujt−1 − ujt, ∀j ∈ G, ∀t ∈ T (2.12)
ζ−i +t
∑
q=t−Dupi
zit 6 uit, ∀i ∈ Gth, ∀t ∈ T (2.13)
ξ−i +t
∑
q=t−Ddni
yit 6 1− uit, ∀i ∈ Gth, ∀t ∈ T (2.14)
− P rlj 6 gjt − gj(t−1) 6 P rl
j , ∀j ∈ G, ∀t ∈ T (2.15)
gjt, rjt > 0, ujt, yjt, zjt ∈ {0, 1}, ∀i ∈ Gw, ∀t ∈ T (2.16)
The power balance constraint (2.2) ensures that in time period t, the sum of power
20
produced by all the generating units connected to the bus k (Gk is the set of all the
generators connected to bus k) is equal to the sum of the demand at the bus k, lkt,
and the summation of power flow in all transmission lines between bus k and other
buses, pkt, minus the unserved demand, luskt . Total power flow out from the bus k
in time t, pkt is calculated by (2.3) using linearized power flow method (also known
as DC power flow) [41]. In (2.3), bkr is the susceptance of the transmission line
between buses k and r. The voltage angles of buses k and r at time t are denoted
by δkt and δrt respectively. Constraint (2.4) ensures that the power flow in a certain
line is within its capacity (±Tkr).
Constraint (2.5) ensures that in time period t, if the unit j is committed, the
sum of its output gjt and the amount of spinning reserves it provides, rjt, is within
the feasible operating range of the unit. Pmaxj and Pmin
j are the maximum and
minimum operating limits of unit j respectively. The commitment variable of unit
j, ujt, determines whether or not the unit is committed in time t.
Constraint (2.6) ensures that the spinning reserves provided by all units satisfy
the spinning reserve demand of the system in time t, Rt. Constraint (2.7) controls
the amount of reserves provided by unit j where, P resj is its reserve limit.
Constraint (2.8) ensures that in time t, the output of each unit in set of gener
ators that have mustrun constraints (Gmr) is equal or greater than the mustrun
limit, Pmrit . The output of each unit in hydro generating unit set Gmr is constrained
by resource availability in time t, P havit (2.9). Similarly, output of wind units is
constrained by wind availability in time t, Pwavit (2.10).
Constraint (2.11) represents the state transition of unit j in time t from off to
on. Similarly, Constraint (2.12) represents the reverse state transition of unit j
in time t where yjt is the binary variable that determines the unit decommitment.
The constraint (2.13) controls the minimum uptime of thermal generating unit Gth
21
where, Dupi is the minimum up time of the thermal unit i. The binary parameter ζ−i
is appropriately set considering the running history of unit i at time t. Similarly,
(2.14) controls the minimum downtime of thermal units where, Ddni is the mini
mum downtime of unit i. The binary parameter ξ−i is set considering, how long
the he unit has been offline prior to time t. Constraints (2.13)(2.14) have been
adopted from [42].
Intertime step ramping of generating units is controlled by the constraint (2.15),
where P rlj is the ramp limit of unit j. Finally, (2.16) ensures the nonnegativity of
decision variables gjtandrjt; integrality of decision variables ujt, zjt, andyjt.
2.3.2 Realtime Operation Model
The objective of the realtime operation model is to dispatch the available generating
units to satisfy the demand in the time period of interest (t∗) while minimizing the
total cost of fuel and cost of unserved demand. The objective function of this model
is given by (2.17) where τ rt is the time step of the model. In this study, τ rt is set to
10 minutes. Similar to the UC model, fj(·) calculates the fuel cost of the unit j.
minimize∑
j∈G
fj(gjt∗) · τrt +
∑
k∈K
C ls · llskt∗ · τrt (2.17)
The objective function (2.17) is subject to constraints (2.18)(2.26). These con
straints are identical to those of the UC model. Power balance is ensured by (2.18)
and (2.19). Transmission limits are enforced by (2.20). Feasible operating region
of a generating unit is controlled by (2.21) while mustrun generation levels are
enforced by (2.22). Hydro and wind generation bounds are set by (2.23) and (2.24)
respectively. The ramping limits are enforced by (2.25). Nonnegativity and inte
grality of appropriate variables are ensured by (2.26). Note that there are no reserve
22
allocations, unit state transitions, and minimum up and downtime limits in the
RO model.
∑
i∈Gk
git∗ = lkt∗ − llskt∗ + pkt∗ , ∀k ∈ K (2.18)
pkt∗ =∑
r∈K
−bkr · (δkt∗ − δrt∗), ∀k ∈ K (2.19)
− Tmaxkr 6 −bkr · (δkt∗ − δrt∗) 6 Tmax
kr , ∀(k, r) ∈ K (2.20)
ujt∗ · Pminj 6 gjt∗ 6 ujt∗ · P
maxj , ∀j ∈ G (2.21)
git∗ > Pmrit∗ , ∀i ∈ Gmr (2.22)
git∗ 6 P havit∗ , ∀i ∈ Gh (2.23)
git∗ 6 Pwavit∗ , ∀i ∈ Gw (2.24)
− P rlj 6 (gjt∗ − gj(t∗−1)) 6 P rl
j , ∀j ∈ G (2.25)
gjt∗ > 0, ujt∗ ∈ {0, 1}, ∀j ∈ G (2.26)
In contrast to UC model, the main difference of the realtime operation model
is that there are no temporal links between decision variables. Therefore, the real
time operation model for different time periods can be run sequentially, rather than
concurrently as in the case of UC model. Unit output of the previous time period
gj(t∗−1) can be supplied as a parameter for the constraint (2.25).
2.3.3 Data and Assumptions
The required data for the simulation study have been obtained from publicly avail
able sources. The generating unit fleet of Alberta in 2008 is used as the baseline
for this study. The choice of 2008 was mainly due to the availability of wind and
demand data at the required resolution (i.e. 10 minutes). These time series data
23
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 16000
7000
8000
9000
10000
Dem
and
(MW
)
Fraction of year
Peak demand = 9835 MW
Baseload = 6384 MW
Average demand = 7963 MW
Figure 2.2: Demand duration curve
sets were obtained from the AESO [47]. The wind power data set is aggregated and
provides the combined output of all the WPPs in Alberta in 2008. That is another
reason to use 2008 data, as the installed wind power capacity in Alberta remained
constant throughout the year. The demand data is also aggregated and provides
the total demand of the system. The aggregated demand is divided among the 6
buses/regions in proportion to the average demand data of each region. The av
erage load fractions have been calculated using the data from [46]. The duration
curve of the total system demand is depicted in figure 2.2.
The generation capacity available at each bus by generating technology is listed
in table 2.1. In 2008 Alberta had 89 power plants. In this model, some of these
power plants are aggregated and represented as a single unit. That includes the
10 WPPs, which are represented as a single unit connected to "South" bus. The
full list of generating units in the model is available in appendix A. Ranges of the
heat rates at rated capacity and ramp rates of the generating units in the model
are summarized in table 2.2.
A key challenge in this study was obtaining the parameters of the generating
units such as their heat rates (i.e. efficiency), start up fuel consumptions, mini
mum up and downtimes, and ramping limits. In power systems with competitive
electricity markets, such information is propriety and kept confidential. Publicly
24
Table 2.1: Generating units available at each bus (in MW)
(1) South (2) Calgary (3) Central (4) Edmonton (5) Northeast (6) Northwest TotalDemand fractiona 0.09 0.20 0.16 0.26 0.18 0.11
Coal 756 0 664 4330 0 143 5893CCGT 0 250 0 0 0 0 250SCGT 27 0 0 46 0 205 278Cogeneration 239 240 575 69 2555 233 3911Hydro 89 320 470 0 0 0 879Biomass 36 0 11 0 0 169 216Wind 497 0 0 0 0 0 497
Total 1644 810 1720 4445 2654 651 11924a Demand at each bus is given as a fraction of the total demand.
Table 2.2: Generating unit heat rates and ramp rates
Generatingtechnology
Heat rate (HHV,GJ/MWh)
Ramp rate(% MCR/min)b
Coal 1216 12SCGT 1216 1015CCGT 8 5Cogen 7.5 25Hydro N/A 20Biomass 12 1a Ramp rate is given as a percentage of themaximum continuous rating (MCR) of a unit.
available sources such as [48] that present parameters of generating units of Al
berta in the prederegulation period (before 2000) are used to obtain the required
data. Estimates have been made for parameters that are not publicly available and
the validity of the estimates have been verified by comparing with published data
such as [49,50]. The most notable parameter estimation is the formulation of fuel
cost functions of coal, SCGT, and CCGT units. A fuel cost function of a thermal
generating unit calculates the fuel consumption at a given output. To formulate the
fuel cost function of coal units, first, the average heat rates are obtained from [48].
It was assumed that for a given unit the published value is the average heat rate
when the unit is operating at its rated capacity. The average heat rate at outputs
bellow the rated capacity is estimated using the estimates of heatrate degradation
of coal power plants available in [51, p.20]. The average heat rate values are used to
estimate the fuel consumption at different outputs. Then a piecewise liner model
is fitted to form the fuel cost function fj(·) of each coal unit [41, p.4950]. Similar
25
method is used to estimate the fuel cost functions of SCGT and CCGT units and
the required data are obtained from [52].
Approximately 30% of the installed generation capacity in Alberta consists of
cogeneration units that satisfy behindthefence industrial demands (see chapter
4 for more details). In the simulation models approximately 5065% of the co
generation unit capacities are set to be must run generation and therefore, not
dispatchable. Cogeneration unit outputs determined by the UC model are fixed for
the RO stage. The heat rates of the cogeneration units are assumed to be constant.
Spinning reserves are also allocated in the scheduling stage using the UC model.
The total spinning reserves requirement (Rt) is assumed to be 5% of the forecasted
demand of a given hour. Only SCGT, CCGT, and hydro units were allowed to
provide spinning reserves.
With respect to the transmission system, the current experiment intends to pro
vide insights of major transmission capacity requirements between regions. There
fore, transmission lines are assumed to have infinite capacities (Tmaxkr ) and infinite
susceptance values (bkr). Consequently, constraints 2.4 and 2.20 are relaxed for
all simulations.
Impacts of wind power penetration levels of 060% (06GW of installed capacity)
are studied in this experiment. In 2008, the installed wind capacity in Alberta was
497MW. The 2008 wind power production data set was linearly scaled to generate
time series datasets (10 minute time steps) at higher capacities. Throughout the
study demand data was kept at same levels as 2008 values.
Hydro resource availability data has been obtained from a previous study of the
Alberta electric system by MacCormack et al. [53].
The wind forecasts required for the UC model are generated by fitting a second
order autoregressive (AR2) time series model to historic hourly average wind power
26
production data using the methods described in [54,55]. The AR2 time series model
is given by (2.27) and it assumes that the wind power availability in a certain hour,
Pwavit , depends on the output of the previous two hours. The model parameters
φ1, φ2, and σw are estimated using the Matlab® System Identification Toolbox.
Pwavit = φ1 · P
wavit−1 + φ2 · P
wavit−2 + ǫt (2.27)
where, ǫt = random normal noise with zero mean and standard deviation σw
Following the approach taken in [54], different sets of model parameters are calcu
lated for each month in a year. The AR2 models and a random number generator
are used to produce wind forecasts with a mean absolute error (MAE) of 15% of
the rated capacity of the WPP. The intention of using a forecast with a consider
able MAE is to simulate the challenges faced by the system operators in operations
planning stages due to the unpredictability of wind and costs that may incur due
to imperfect knowledge wind. At the operations planning stages, there is uncer
tainty in system demand as well. However, as the focus of this study is limited to
wind power, perfect foreknowledge of demand at the planning stage is assumed.
Consequently, the hourly average of the actual demand data set is used for the UC
model simulations.
The models calculate the fuel consumption of each generating unit in each
time period. The CO2 emissions are calculated by multiplying the fuel amount by
respective fuel carbon intensity. The prices and CO2 intensities of different fuels
are listed in table 2.3. Nonfuel operating costs of all units are assumed to be
negligible. The cost of unserved demand is set to be 1000 $/MWh. All cost values
are in 2010 Canadian dollars (CA$(2010) 1 = US$(2010) 0.97).
27
Table 2.3: Fuel prices and carbon intensities
Fuel type Price ($/GJ) Carbon intensity (tCO2/GJ)
Coal 1 0.1a
Natural Gas 4 0.05a
Biomass 0 0Wind, hydro 0 0a Source: [56].
2.3.4 Model Implementation and Simulation Workflow
The UC model and the RO model are formulated as mixed integer programming (MIP)
problems. The two models are implemented in Matlab®/Tomlab® environment and
solved using CPLEX 12.1® solver. The piecewise linear fuel cost function, fj(·), is
implemented using builtin subroutines in the Tomlab optimization environment.
The simulation workflow for a period of a single day is described bellow:
Step 1 Generating units for each of 24 hours are scheduled by running the UC
model considering the forecasted wind power availability. Generating units
that provide spinning reserves are also selected.
Step 2 RO model is run for each of 144 ten minutes time periods of the day.
Commitment of slow start thermal units (i.e. coal, CCGT, cogeneration, and
biomass) were fixed to the schedules set by the UC model (ie. only the units
that have been scheduled for a certain hour by the UC model are available to
satisfy the demand in 6 ten minutes periods within that hour). Unscheduled
startups are allowed only for SCGT and hydro units. Units that are selected
to provide reserves in step 1 are required to operate at least at their mini
mum operating level. Output levels of the cogeneration units are fixed to the
amounts determined at step 1.
Step 3 Results are tabulated and parameters required to simulate operations of
the following day (such as running history, output level of the final time step
28
etc.) are carried forward for the simulations of the next 24 hours period.
Steps 13 are repeated for 365 days at each wind penetration level. A total
of 8 wind penetration levels in the range of 060% have been simulated in this
experiment. A single year has 365 UC model runs and 52,560 RO model runs. The
average time required to simulate the operations of a single year at a given wind
penetration level on a Mac OS X 10.6® based computer running at 2.26GHz with
4GB of RAM is approximately 90 minutes.
2.4 Results and Discussions
Simulation results at each wind penetration level are used to evaluate power system
operating costs, CO2 emissions, impacts on other generating units, and transmis
sion system requirements.
The total electricity demand in the simulated year is 69,800 MWh. Shares of
electricity supplied by each generation technology to satisfy that demand at differ
ent amounts of installed wind capacity are depicted in figure 2.3. It can be seen
from this figure that as wind penetration level increases, wind power competes with
coal generated electricity and displaces that in a fully competitive market setting,
where generating units compete with their marginal costs. As wind penetration
increases from 0% to 60%, the share of coal reduces from 66% to 41%. One caveat
of these results is that 5060% of the natural gas fired cogeneration capacity is
constrained to be mustrun and therefore, that energy volume is dispatched all the
time. In case of Alberta, this is not unreasonable. As discussed in chapter 4, co
generation units follow the host facility’s thermal energy demand and offer energy
to the market at zero dollars. Furthermore, it is reasonable to assume that majority
of the behind the fence industrial electricity demand is satisfied by onsite cogen
eration units. The share of dispatchable natural gas reduces from 1% at the no
29
0 5 10 20 30 40 50 60
Windpower generation capacity (% peak demand)
0 500 1000 2000 3000 4000 5000 60000
10
20
30
40
50
60
70
80
90
100
Wind power generation capacity (MW)
Per
cent
age
of to
tal d
eman
d(%
)
CoalCCGTSCGTCogenHydroBiomassWind
Figure 2.3: Electricity produced by different generation technologies.The total electricity demand of the simulated year is 69,800MWh. The peak demand ofthe year is 9,835MW (winter peak). Approximately 26% of the demand is satisfied bycogeneration units that operate as must run capacity since they follow the thermal demandof corresponding host facilities. Dispatchable gas share (SCGT, CCGT, and cogenerationbeyond must run capacity level) increases after 20% wind penetration level in order toprovide flexible capacity to firm wind power.
30
0 5 10 15 20 25
Wind power generation (% total demand)
0 10 20 30 40 50 60 700.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
CO
2 em
issi
ons
inte
nsity
(tC
O2/
MW
h)
Wind power generation capacity (% peak demand)
All units (EI 1)All units excluding wind (EI 2)All dispatchable units (EI 3)
EI 2
EI 3
EI 1
Figure 2.4: Average CO2 intensity of the energy mix.Average CO2 intensity of the energy mix is calculated considering emissions from consumedboth at unit startups and electricity producing stages. Line EI 1 depicts the averageCO2 intensity of the total electrical energy mix. Line EI 2 depicts the CO2 intensity ofelectricity from all units except wind. Line EI 3 depicts the CO2 intensity of electricityfrom all dispatchable units that are competing with wind (ie. excludes wind and mustruncogeneration).
wind case to 0.5% at 20% penetration, and increases from there, mainly to provide
sufficient ramping capability. This is examined in detail in section 2.4.2.
2.4.1 Carbon Abatement Cost
Average CO2 intensity of the electrical energy mix as a function of the wind power
penetration level is depicted in figure 2.4. The figure also depicts the CO2 inten
sity of the electricity from the units that compete with wind power, by estimating
the CO2 intensity of electricity from all dispatchable units (ie. coal, SCGT, CCGT,
31
cogeneration beyond must run capacity, hydro, and biomass). The results provide
interesting insights into the generation units displaced by wind power. When wind
power is introduced to the system (0% to 5% penetration), the CO2 intensity of
dispatchable energy mix marginally increases because natural gas fired electricity
at the margin is displaced by wind. From 5% to 40% penetration level, wind power
displaces the coal units that have higher marginal costs due to lower efficiency (or
high heat rate). Therefore, within that range the CO2 intensity of the dispatchable
energy mix decreases. From there, the CO2 intensity rises marginally because ma
jority of the coal units operates at lower operating points, thus with low efficiencies
(This fact is reexamined in section 2.4.2). The higher number of units startups
(mainly SCGT units) is another a contributing factor.
Figure 2.5 shows the average operating costs incurred in satisfying the elec
tricity demand with the mix of generating units in the model (both startup and
operating fuel costs are considered). As there is no fuel cost in wind power gen
eration, the average fuel cost reduces with increasing wind penetration level. Two
other cost scenarios are depicted in figure 2.5. In cost scenario S1, the capital
costs of wind are added to the operating costs. In scenario S2, both the capital
cost of wind and that of new SCGT capacity required to provide sufficient system
flexibility are added to operating costs. The new SCGT capacity requirements are
iteratively determined so that the cumulative load shedding time is less than 1%
of the total time. New SCGT capacity required at each wind penetration level are
listed in table 2.4. Capital costs of wind and SCGT are assumed to be $2500/kW
and $1000/kW respectively (both are in 2010 Canadian dollars). These values are
obtained from the most recent estimates made by the AESO [44]. Capital costs are
amortized over 20 years using a discount rate of 12%.
Results depicted in figure 2.5 are then combined to estimate the carbon abate
32
0 5 10 15 20 25Wind power generation (% total demand)
0 10 20 30 40 50 60 700.5
0.6
0.7
0.8
CO
2 em
issi
ons
inte
nsity
(tC
O2/
MW
h)
0 10 20 30 40 50 60 700
20
40
60
Ave
rage
cos
t of e
lect
ricity
($/
MW
h)
Wind power generation capacity (% peak demand)
Average fuel costAverage cost (S1)Average cost (S2)Carbon intensity
Figure 2.5: Carbon emissions intensity and average cost of electricityThree electricity cost scenarios are depicted in this figure. The ‘‘Average fuel cost" curve isthe average of the fuel consumed at unit startup and electricity generation stages. ScenarioS1 is the sum of total fuel costs and capital costs of wind power. Scenario ‘‘S2" is the sumof total fuel costs, capital costs of wind power, and capital costs of new SCGT capacity.The CO2 intensity curve is that of the total energy mix. Capital cost of wind and SCGT areassumed to be $2500/kW and $1000/kW respectively [44]. A Discount rate of 12% over20 years period is used to amortize the capital costs.
ment cost of wind power. The carbon management option investigated in this
analysis is adding wind power to the existing power system. In order to estimate
the CO2 abatement cost, a baseline has to be chosen. Power generating unit fleet
in Alberta in 2008 is assumed to be the baseline for this analysis. The marginal
abatement cost is defined as the ratio between the total mitigation cost and the
abated emissions. Total mitigation cost at a certain wind penetration level is the
sum of the capital costs incurred in adding wind power and the operating cost
(ie. fuel cost) differential relative to the baseline. Figure 2.6 depicts the relative
33
Table 2.4: New SCGT capacity required at different wind penetration levels
Installed wind capacity (MW) 0 500 1000 2000 3000 4000 5000 6000
Wind penetration level (%) 0 5 10 20 30 40 50 60New SCGT capacity (MW) 0 0 200 500 700 1000 1000 1500
CO2 emissions reductions potentially achieved by adding wind power into the Al
berta electric system and the corresponding abatement costs (This type of results
are know as carbon abatement supply curves). Marginal abatement costs are cal
culated under both S1 (operating cost + wind capital cost) and S2 (operating cost
+ wind capital cost + new SCGT capital cost) cost scenarios.
As shown in figure 2.4, the emissions reduction potential is proportional to the
installed wind capacity. For example, 216 million tonnes of CO2 can be abated per
year by increasing the wind penetration level to 1060%. However, the marginal
cost increases at higher abatement levels. This is due to ancillary costs and emis
sions from the measure to firm wind variability. Even after accounting for the
capital costs of new capacity and variability cost, wind power provides an effective
option to decarbonize the Alberta electric system. For example, at 40% penetra
tion level, wind power can potentially avoid 11 MtCO2 of emissions per year at an
abatement cost of 109 $/tCO2 (under cost scenario S2). The abated amount is
approximately 5% of Alberta’s total CO2 emissions in 2008 [45]. Furthermore, at
the same penetration level, the CO2 intensity of the electricity mix is reduced by
20%.
The marginal abatement cost of wind power is competitive compared to carbon
capture and storage (CCS). CCS is a mitigation option that is being promoted and
government mandated [57] in Alberta as a climate change mitigation option. The
marginal abatement cost of CCS in Alberta is estimated to be 92122 $/tCO25
for new coal fired facilities to abate 716 MtCO2/year [58]. Abatement costs of
5The cost figure given in [58] is converted to 2010 Canadian dollars by adjusting for inflation
34
0 2 4 6 8 10 12 14 16
Abated carbon emissions (MtCO2/year)
0 5 10 15 20 25 3090
95
100
105
110
115
120
Percentage carbon emissions abatement (Base wind capacity = 500MW)
Cos
t of c
arbo
n ab
atem
ent (
$/tC
O2)
Cost scenario: S1Cost scenario: S2
Figure 2.6: CO2 abatement supply curve.This figure depicts the marginal CO2 abatement cost of wind power in Alberta under thecost scenarios S1 and S2 (Cost scenario S1= fuel cost + capital cost of wind; scenario S2 =fuel cost + capital cost of wind + capital cost of new SCGT capacity). The ‘‘baseline" windcapacity is 500MW (5% penetration level). At higher CO2 abatement levels the marginalabatement cost rises due to increasing costs and emissions associated with firming windpower plant outputs.
retrofitting existing power plants with CCS are estimated to be higher at 163
255 $/tCO2. Therefore, we conclude that wind power has a competitive marginal
abatement cost compared to CCS.
Marginal abatement costs calculated in this section are sensitive to the cap
ital costs of wind and SCGT facilities. The values used for this analysis (wind:
$2500/kW; SCGT: $1000/kW) are from the AESO estimates and manifest the high
engineering and construction costs in Alberta. Assuming the U.S. Energy Infor
mation Administration’s estimates for the capital costs (wind: $2100/kW; SCGT:
35
$750/kW) would reduce the marginal abatement costs by approximately 20% in
both S1 and S2 cost scenarios.
2.4.2 Cost of Wind Variability
In this section, two aspects of wind power—uncertainty and variability—that can
lead to higher costs and emissions are examined. The uncertainty of wind affects
planning and operations of a power system because the system operators can not
forecast the availability of wind in a certain point of time in the future with 100%
certainty . The MAE in a WPP output forecast made 24 hours ahead can be as
high as 2040% of the capacity [25]. Therefore, generating unit scheduling and
reserve procuring are made with imperfect information. System impacts due to
differences between the forecasted and actual wind power production are mitigated
through ancillary services and energy market dispatches that may be out of merit.
Similarly, variations of the WPP outputs should be firmed by changing the output
of the dispatchable units. In extreme situations, demand may have to be curtailed,
adding significant operating costs.
In this study, the cost of uncertainty and variability is defined as the difference
between the total operating cost calculated by the UC model and that by the RO
model. The UC model schedules the generating units with a wind forecast and
also assumes constant WPP outputs within an hour. The RO model, which runs
at 10 minutes time steps, captures the actual intrahour variability of wind. Any
shortages in available generation capacity relative to the committed capacity are
procured by committing a fast start unit (SCGT, hydro). Therefore, the operating
cost differential captures the impacts of both uncertainty and variability of wind. It
is assumed that generating units for a certain day are scheduled at 12:00 hours of
the previous day. This is analogous to dayahead energy market run by a system
operator. All assumptions are identical to the ones described in section 2.3.3.
36
11.5
12
12.5
13
13.5
14
14.5
15
15.5
Ave
rage
ope
ratin
g co
st (
$/M
Wh)
0 10 20 30 40 50 60 70800
850
900
950
1000
1050
1100
Tot
al o
pera
ting
cost
(m
illio
ns $
/yea
r)
Wind power generation capacity (% peak demand)
Day−ahead scheduling (C1)Real−time operation (C1)Day−ahead scheduling (C2)Real−time operation (C2)
(a)
0
1
2
3
4
Win
d v
ariabili
ty c
ost
(% a
vera
ge w
ind c
ost)
0 10 20 30 40 50 60 700
1
2
3
4
5
Wind power generation capacity (% peak demand)
Win
d v
ariabili
tycost
($/M
Wh o
f w
ind e
nerg
y)
C1C2
cost
($/M
Wh o
f to
tal energ
y) (b)
Figure 2.7: Cost of wind uncertainty and variabilityFigure 2.7(a) shows the total operating costs estimated at the generating unit schedulingstage using the UC model (Dayahead scheduling) and the operating costs incurred in realtime operation. The curves denoted as ‘‘C1’’ correspond to total operating costs when dayahead unit scheduling is done with a wind forecast with a mean absolute error equivalentto 15% of the installed wind capacity. In contrast, curves denoted as ‘‘C2’’ correspond tothe operating costs when dayahead unit scheduling is done with perfect foreknowledge ofhourly wind availability in the planning period. Figure 2.7(b) depicts the cost of uncertaintyand variability per MWh of wind energy. The results suggest that in the power systemstudied, the cost of wind uncertainty and variability is moderate even at very high windpenetration levels.
37
In figure 2.7a, the two curves denoted as ‘‘C1’’ depict the annual operating
costs calculated by the UC model (‘‘Dayahead scheduling (C1)’’) and RO model
(‘‘Realtime operation (C1)’’) under each wind penetration level (these results are
henceforth referred to as ‘‘case C1’’). As explained in section 2.3.3, wind forecast
used for UC model has a MAE of 15% of the total installed wind capacity. It is
evident that there is a considerable difference between the operating costs estimated
at dayahead scheduling stage and operating costs incurred in realtime operation.
The uncertainty and variability of wind power increases the total operating cost by
0.28% at wind penetration levels of 560%.
Another simulation experiment is carried out to estimate the cost variability
alone (this experiment is henceforth referred to as the ‘‘case C2’’). To do so, first
a wind forecast with perfect knowledge of ‘‘hourly wind availability’’ is formulated
by moving averaging the 10 minute data set with a hour long window. Then, the
operations of the full year are simulated again as described in section 2.3.4 using
the perfect forecast at Step 1 (UC model runs). Since the operator scheduled the
generating units with perfect foreknowledge of hourly wind energy availability the
operating cost difference between the UC model and RO model can be considered
as the cost of wind variability. The annual operating costs calculated by this
experiment are depicted by the curves denoted as ‘‘C2’’ in figure 2.7a.
Cost of uncertainty and variability per MWh of wind power is depicted in fig
ure 2.7b6. As shown in the figure, the cost of uncertainty and variability at 5% pen
etration level is approximately $1/MWh of wind energy and increases to $4/MWh at
60% penetration level. Adding the uncertainty and variability cost would increase
the average cost of wind energy by 13.5%.
Comparing the results under the cases C1 and C2 shows that the cost of un
certainty is more significant compared to the cost of variability. That affirms the
6Cost of uncertainty & variability =RO operating cost−UC operating cost
Wind energy production
38
importance of a reliable wind power forecasting for operations planning of a power
system with significant wind capacity.
Similar to the cost, the increase in CO2 emissions due to uncertainty and vari
ability is calculated by subtracting emissions estimated at dayahead scheduling
from the emissions of realtime operation. Figure 2.8 shows the total CO2 emis
sions in the simulated year (top figure) and the average CO2 emissions increase per
MWh of wind energy (bottom figure). It can be seen that, depending on the wind
penetration level, the ancillary CO2 emissions associated with measures to mitigate
the uncertainty and variability of wind power amounts to 0.040.09 tCO2/MWh of
wind energy. These ancillary costs and emissions contribute to the higher marginal
CO2 abatement costs at higher wind penetration levels (Figure 2.6).
The relatively lower cost of uncertainty and variability of wind is a surprising
result. This result suggests that, even with lower ramping ability, coal units that
have relatively lower marginal cost can follow wind variations and provide firming
power (Ramp rates of 17 of the 18 coal units in the models are in the range of 23
MW/minute and the remaining unit has a ramp rate of 5 MW/minute). To warrant
this assertion and to gain insights into the firming power providers, a correlation
coefficient analysis has been carried out. First, time series data sets of intertime
step ramps of net demands7 at a given wind penetration level, and outputs of coal
fired units, natural gas fired units, and hydro units (coal, gas, and hydro units
are appropriately aggregated to form three technology groups) are formulated by
aggregating and differencing (biomass units are excluded because of their very low
ramp rates and smaller capacity). Then, the correlation coefficients between the
intertime step ramping of net demand and other dispatchable generating units (ie.
coal, natural gas, and hydro) are calculated. Similarly, the correlation coefficients
between the ramp time series data sets of wind and other dispatchable generat
7net demand= demand wind
39
0.55
0.6
0.65
0.7
0.75
0.8
0.85
Ave
rage
CO
2 em
issi
ons
(tC
O2/
MW
h)
0 10 20 30 40 50 60 7035
40
45
50
55
60
Tot
al C
O2 e
mis
sion
s (M
tCO
2/ye
ar)
Wind power generation capacity (% peak demand)
Day−ahead scheduling (C1)Real−time operation (C1)
0 10 20 30 40 50 60 700.02
0.04
0.06
0.08
0.1
Wind power generation capacity (% peak demand)
Ave
rage
CO
2 em
isis
ons
(tC
O2/
MW
h of
win
d en
ergy
)
Figure 2.8: CO2 emissions stem from mitigating uncertainty and variability of windpower.The top figure shows the total CO2 emissions estimated at the dayahead unit schedulingstage and the CO2 emissions in realtime operation. Dayahead unit scheduling is donewith a wind forecast that has a mean absolute error equivalent to 15% of the installed windcapacity (case ‘‘C1’’). The bottom figure shows the CO2 emissions resulting from mitigatinguncertainty and variability of wind power in tCO2 per MWh of wind energy.
ing units are calculated. These results are depicted in figures 2.9a and 2.9b. A
strong correlation between the ramps of net demand and coal and a strong negative
correlation between wind and coal at all wind penetration levels can be observed
from these figures. This confirms the fact that coal units follow the wind variations
(From here onwards the discussion is limited to correlations between ramps of wind
and other generating units for simplicity). However, at higher penetration levels,
the negative correlation between wind ramps and coal ramps declines. This is due
40
0 10 20 30 40 50 60 700
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Wind power generation capacity (% peak demand)
Cor
rela
tion
Coe
ffici
ent
Net demand/CoalNet demand/Natural GasNet demand/Hydro
(a)
0 10 20 30 40 50 60 70−1
−0.9
−0.8
−0.7
−0.6
−0.5
−0.4
−0.3
−0.2
−0.1
0
Wind power generation capacity (% peak demand)
Cor
rela
tion
Coe
ffici
ent
Wind/CoalWind/Natural GasWind/Hydro
(b)
Figure 2.9: Correlation between intertime step ramps of WPPs and other dispatchable generating unitsFigure 2.9(a) depicts the correlation coefficients between the inter time step (in 10 minutesteps) ramps of net demand (net demand = demandwind power production) and otherdispatchable generating units (ie. coal, gas, and hydro). Figure 2.9(b) depicts similarresults for ramps of wind alone. A strong negative correction between ramps of wind andcoal suggests that coal units follow the wind power variations.
41
to the greater wind ramp magnitudes compared to the aggregated ramping ability
of the available coal units. The negative correlation between wind and gas/hydro
units is weak. However, the negative correlation increases at higher penetration
levels, manifesting gas and hydro units’ firming of wind variations.
As explained above, the coal units in the model provide the firming power at
all penetration levels, lowering the variability cost of wind. However, in order to
firm wind coal units will have to be redispatched more frequently. Such cyclical
operations of coal units can lead to higher operating and maintenance costs8. Fig
ure 2.10 displays the operating level of the 18 coal units as a percentage of MCR at
different wind penetration levels. It is evident form the figure that as the amount of
wind in the system rises, operations of the coal units become increasingly cyclical.
Thermal generating units, particularly coal fired units, are not optimal for cyclical
operations. As discussed in [27, 59–61] such cyclical operations have significant
impacts on generating unit equipments and can lead to failures, higher forced
outage rates, and consequently higher maintenance cost. It is difficult to make a
general estimate of the increase in maintenance costs of coal power plants due to
the diversity in plant designs and has to be done plant by plant basis. Such an
estimate is not undertaken in this study due to the unavailability of data and left
for future work. However, through empirical analysis in [59] Lefton & Besurner
find that older coal power plants such as the ones in Alberta are less susceptible
to damages due to cyclic operations. Hence, the increase in operating and main
tenance costs of coal units due to cyclical operations might not significantly alter
the results of the study.
8The term ‘‘cyclical operation’’ is refereed to startup/shutdown operation, onload cycling, andhigh frequent MW changes [59]
42
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 0MW%
of M
CR
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 500MW
% o
f MC
R
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 1000MW
% o
f MC
R
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 2000MW
% o
f MC
R
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 3000MW
% o
f MC
R
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 4000MW%
of M
CR
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 5000MW
% o
f MC
R
0
0.5
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Unit ID
Installed Wind Capacity= 6000MW
% o
f MC
R
Figure 2.10: Coal unit operationsThis set of figures shows the ‘‘boxplots’’ of operating levels of 18 coal fired generating unitsin the simulated year under different wind penetration levels. The operating level of a givenunit is depicted as a percentage of its MCR. The top and bottom edges of a box denote the75th and 25th percentile of the operating level. The red coloured horizontal line denotesthe median operating level and the edges of the whiskers corresponds and upper and loweradjacent values. Outliers are denoted by ‘‘+’’ markers. Unit 6 is a supercritical unit andhas the highest efficiency. The minimum stable operating level is 40% of MCR. All othersare subcritical units and the minimum stable operating level of each of them is 35% ofMCR.
43
0 20 40 60 80 1000
1000
2000
3000
Line: 1 2
Pow
er
flow
(M
W)
Duartion (% of time)0 20 40 60 80 100
0
1000
2000
3000
Line: 1 3
Pow
er
flow
(M
W)
Duartion (% of time)
0 20 40 60 80 1000
500
1000
1500
Line: 2 3
Pow
er
flow
(M
W)
Duartion (% of time)0 20 40 60 80 100
0
1000
2000
3000
Line: 3 4
Pow
er
flow
(M
W)
Duartion (% of time)
0 20 40 60 80 1000
500
1000
Line: 4 5
Pow
er
flow
(M
W)
Duartion (% of time)0 20 40 60 80 100
0
500
1000
1500
Line: 4 6
Pow
er
flow
(M
W)
Duartion (% of time)
0 20 40 60 80 1000
200
400
600
800
Line: 5 6
Pow
er
flow
(M
W)
Duartion (% of time)
500 MW
1000 MW
2000 MW
4000 MW
5000 MW
6000 MW
Installed wind capacity
Figure 2.11: Power flow duration curvesThis set of figures show the duration curves of magnitudes of power flows in the seventransmission lines in the models at different wind penetration levels. The numbers insideeach subfigure denote the two busses connected by the corresponding transmission line.See figure 2.1 for the configuration of the buses and transmission lines.
44
2.4.3 Transmission System Impacts
Another challenge to integrate large amounts of wind power into existing power
systems is the need for new transmission infrastructure (that includes both new
transmission lines and reinforcing the existing transmission lines). The transmis
sion lines in the simulation models represent the major transmission corridors in
Alberta (See figure 2.1). In this section, the major transmission requirements to
support largescale wind power in the province are investigated. Figure 2.11 de
picts the duration curves of the magnitudes of MW flows in the seven transmission
lines at different wind penetration levels. Discussion in this section is limited to
an examination of MW flow changes in the 7 transmission lines relative to the
baseline wind capacity (500MW). Cross checking with corresponding historic MW
flows in lines 23, 34, and 46 verified that the model calculated flow durations
are satisfactorily close to actual values.
Examining the peak MW flow in different lines show that capacity requirement
of line 12 (between ‘‘South’’ and ‘‘Calgary’’) and line 13 (between ‘‘South’’ and
‘‘Central’’) are comparable and proportional to the installed wind capacity. The ca
pacity of the line 23 remains unchanged up to 60% of wind penetration level. Only
marginal changes (less than 100MW) in peak transmission requirements are ob
served in all other lines up to wind penetration level of 40% (serving approximately
16% of the total demand).
Transmission cost in adding new wind capacity is not estimated in this study
and left for future work. However, as estimated in previous studies such as [36],
the transmission cost is relatively small and may not significantly alter the results
of the study.
45
2.4.4 Caveats and Limitations of the Study
Electric power systems are large and complex engineering systems and a model
ing exercise such as this work is done by reducing that complexity through set
of assumptions to retain simplicity, mathematical tractability, and clarity of the
discussion. Although, maximum care has taken to minimize the impact of such
assumptions, they inevitably affect the results. Nevertheless, we are confident that
the final results are in the correct order of magnitude compared to results produced
by complete representation of the actual system. Some major caveats and limita
tions of the model are discussed here and directions to improve them in future
modeling exercises are provided.
One of the most significant challenges for this modeling exercise was finding the
heat rate values of thermal generating units. They are at the centre of the model
and significantly influence the final results. The values used for these models
are from the very few publicly available sources. Future research should strive
to minimize the uncertainty in those values. With respect to wind power studies,
ramp rates of conventional units are another important set of parameters. Very
conservative values that are based on published sources have been used for this
study.
The dispatch models developed for this study represent a fully competitive elec
tricity market, where electric power generators compete with marginal costs. How
ever in actual deregulated electricity markets some units do have market power and
the energy bids do not necessarily reflect the marginal costs of the corresponding
units. Studying the competitive behaviour of generating units and its impact on
wind power’s effectiveness for carbon management is an interesting area for future
research. But it is beyond the scope of this work.
In the current model, 100% availability is assumed for all the generating units.
46
However, generating units have both planned and unplanned outages. Additional
parameters that are estimated using probabilistic methods can be introduced to
the models to represent generating unit outages.
The models do not include interconnections to neighbouring power systems. In
Alberta, the current capacity of the major interconnection line is limited (600MW)
and the import and export volumes are relatively small. Nevertheless, keeping in
terconnections at scheduled flow rates is an important reliability obligation and
therefore should be included in future enhancements of the models. Interconnec
tions can be modeled as price sensitive generators or by using historic flow rates.
Crosschecking with historic generation volumes obtained from [62] shows that
compared to the 2008 data, where the wind capacity was 500MW, the model es
timated coal fired generation volume is within 5% of the actual value. The total
natural gas fired generation volume is within 11% of actual volume. The error in
hydro generation was very significant at about 20%. However, the erroneous vol
ume was very small (less than 1%) compared to total generation. In case of natural
gas fired generating units, future developments to the model should improve the
representation of cogeneration units. In this work, they are represented as units
with constant must run capacity obligations. Future implementations can still take
the same approach; but should focus more on determining the amounts of must
run capacities.
Current models do not capture wind variability in seconds to 10 minutes time
frames. If sufficient wind power and demand data is available, impacts in that
time frame such as increased reserve requirements can be best estimated using
probabilistic methods such as the ones presented in [34,63]. Results of such an
analysis can be used to set reserve requirement limits of the models (ie. constraint
(2.6)).
47
2.5 Conclusions
Reducing CO2 emissions from the electric power sector is an important climate
change mitigation option. Wind power is a proven technology that can be used as a
nearterm option to produce electricity without CO2 emissions. The research work
presented in this chapter evaluated the effectiveness of largescale wind power for
carbon management of electric power systems. Operations of the electric power
system of the Canadian province of Alberta at high wind penetration levels are
simulated using a model that has sufficient resolution to capture the wind power
dynamics.
It was shown that in Alberta wind power can abate 216 million tonnes of
CO2 emissions per year at a marginal abatement cost in the order of 110120
$/tCO2. At an aggressive 60% penetration level, serving a quarter of the total
electricity demand, wind power can reduce the carbon intensity of the Alberta
electric system by 30% compared to 2008 levels.
In the power system studied in this work, the cost of mitigating the system
operational challenges due to the uncertainty and variability of wind power is a
modest 14 $/MWh of wind power at wind power penetration level of 560%, serving
225% of the total demand. Furthermore, the CO2 emissions associated with the
measures to overcome those challenges amounts to 0.050.09 tCO2/MWh of wind
energy.
A general conclusion that can be drawn from this work is that a significant
amount of wind power can be added to a relatively inflexible power system without
drastic increase in cost of wind variability. The vast majority of the dispatchable
units in Alberta are coal fired units. But the results suggest that the cost of
variability and uncertainty of wind, even while serving a quarter of the demand, is
moderate.
48
The analysis and results presented in this chapter are intended to inform cli
mate change mitigation policy makers. The climate change mitigation option we
investigated and proven to be effective is producing electricity without carbon emis
sions by integrating a large amount of wind power into an existing power system in
the near term where majority of the system is remained in place. Policy makers can
use the carbon abatement supply curves developed in this work to compare this
option with other competing and supplementing carbon management options. The
emissions reduction levels and marginal abatement costs presented in this work
are estimated relative to a baseline generating unit mix. Future changes of the gen
erating unit mix can alter the results. The results of this work can be improved by
modeling periods of multiple years where significant changes occur in the system
such as generating unit retirements and new unit additions.
Finally, it can be concluded that large scale wind power, even with accounting
for capital costs and variability cost, is an effective option for carbon management
of the electric power system of Alberta.
49
Chapter 3
Risk Averse Shortterm Operations Optimization of Wind
Power and Compressed Air Energy Storage Systems
3.1 Introduction
The rapid growth of wind power over last decade or so has led to an increasing share
of wind power in the mix of installed power generation capacity in a number of ju
risdictions. Some notable examples include Denmark, Portugal, Spain, Germany,
Texas, Colorado, Iowa, and Minnesota [64–66]. Favourable public policy options
such as production credits, fixed feedin tariffs, and portfolio standards adopted
by many jurisdictions have facilitated the growth of wind power [17,19,67–69]. As
the wind power industry matures, relying on subsidies may not be sustainable and
wind power producers would prefer to maximize their profits by participating in
electricity markets. Furthermore, as wind starts to supply a significant share of
power generation, wind power plants (WPP) may be required to comply with similar
market rules as applied to other generators [70, 71]. However, conventional elec
tricity markets, such as the dayahead market (DAM) are designed for dispatchable
generation. This chapter reviews the challenges for wind power producers partici
pating in a DAM and explores a strategic solution to mitigate those challenges.
3.1.1 Dayahead Market
A DAM for electricity is operated by many competitive power markets to ensure
system reliability through a market based approach. All DAMs are run as auctions
by a market operator and have three main steps [72]: i) bids (pricequantity pairs
50
for given time intervals that typically are 1 hour long) to sell and to buy electricity for
a set period of time are submitted; ii) some bids are accepted through the auction
model, subject to the physical constraints of the power system; the market clearing
price is determined (market clearing price, in general, is the intersection between
the aggregated supply and demand function); iii) accepted bids are settled at the
market clearence price. The cleared bids are financially binding and deviation from
the cleared bids (henceforth called as ‘‘imbalances’’) have to be settled according
to the market rules. Some notable DAMs can be found in PJM Interconnection1,
California, New York, Scandinavia (Nord Pool), Spain, Ontario, and Australia [70,
71, 73]. A significant fraction, if not all, of installed power generation units in a
particular power system are required to bid in to DAM by market rules to provide
enhanced planning flexibility to the transmission system operator (TSO). WPPs in
many jurisdictions currently have the option to participate or not to participate
in DAM [70, 71]. Due to the natural characteristics of wind power, WPPs face
challenges in participating in a DAM.
While the exact market rules depend on how a particular DAM is organized,
in general the market participants are required to submit the bids to buy or sell
electricity for a 24 hour period well in advance of actual delivery hour. For example,
in PJM and Nord Pool, bids for 24 hour period must be submitted up to 12:00 of the
previous day, while in Spain, the market closure is at 10:00. The output of a WPP is
variable combined with considerable uncertainty in the availability of wind power.
Therefore, to participate in a DAM, a WPP has to forecast the wind power generation
to determine the bids. Wind power forecasting has advanced considerably and
very sophisticated computational tools are employed by wind power producers and
TSOs (see [74] for a review of wind forecasting methods). Depending on the market
1A regional transmission organization that operates a transmission system, serving a number ofeastern states in the Unites States
51
rules, wind forecasting has to be carried out as early as 38 hours ahead of the
actual delivery time and at that time frame, making perfect forecasts of wind power
availability is not possible [25,74,75]. The error in a forecast made that far ahead
can be as high as 2030% of the total capacity of the WPP [25,74]. Since bids that
were cleared in the DAM auction are financially binding, a participating WPP is at
the risk of incurring imbalance penalties. Settlement of energy imbalances depends
on the respective power market. In some cases, the imbalances are settled at the
market price while in other cases, imbalance penalties mat be charged [71,76–78].
Another challenge for WPP is the lower economic value of wind power compared
to other conventional generators. In an electricity market, different generators are
dispatched in a least cost approach and settlement is done at the market clearing
price. In a perfectly competitive situation, the market clearing price in a certain
point of time, in general, is equal to the variable cost of the marginal generator [72].
Therefore, the market price and the demand are correlated. Wind power availability
may be weakly or negatively correlated with the demand, and consequently with
the system price, reducing the economic value of wind power. Market conditions
that may lead to lower economic value of wind power has been studied recently
in [79,80].
A number of solutions has been proposed to mitigate the aforementioned chal
lenges. Flexible market rules have been studied in [81–83] and the use of various
forecasting strategies and their optimal integration into bidding decision making
are studied in [84, 85]. Coordinated bidding strategy with a hydro generator is
proposed in [85]. A comparison of financial hedging strategies and physical hedg
ing strategies to lower the risks associated with trading wind energy is presented
in [86]. A strategic bidding method that minimizes the expected imbalance cost is
proposed in [87] and a technique to formulate optimal offerings to a market with
52
different trading floors is proposed in [88].
Another solution is the integration of wind generation with large scale electric
energy storage (EES) systems such as pumped hydro storage (PHS), advanced bat
tery banks, and compressed air energy storage (CAES) systems [89–91]. Reviews of
EES technologies are presented in [91–93]. EES systems may be used to manage
energy imbalances and to time shift wind energy into high demand periods. The
value of EES systems such as CAES for power system applications, including their
feasibility to mitigate wind integration challenges, has been studied in recent litera
ture. Energy arbitrage value of EES is studied in [94–97] and their ancillary service
value is studied in [95]. The broader economics and competitiveness of CAES sys
tems for wind power firming are studied in [23,35,90,98,99]. The economics of
colocating wind and CAES to increase transmission line utilization and decrease
transmission cost is studied in [100]. Use of EES to increase the economic value
of wind power is studied in [79]. A optimal operations strategy for a CAES system
is presented in [101]. A stochastic operations optimization model for a PHS system
and wind generator that are operating either in standalone or jointly is presented
in [76]. This chapter focuses on the use of CAES to mitigate the challenges pose
by the intermittency of wind.
3.1.2 Compressed Air Energy Storage Systems
CAES, as depicted in figure 3.1, is a variation of combustion turbine based elec
tricity generation, where compression and expansion are shifted in time. Air is
electrically compressed and stored, usually underground in solution mined salt
cavities, in mined hard rock cavities, or in aquifers. During the expansion phase
the fuel, usually natural gas, is combusted in pressurized air withdrawn from the
storage. The natural gas requirement is significantly lower ( 60%) compared to a
conventional combustion turbine as no fuel is consumed for air compression. A
53
Figure 3.1: CAES system configurationFigure source: [102]
detailed review of CAES technology is presented in [92,102]. Two CAES systems
are presently in commercial operation, one in Huntorf, Germany, which has been
in operation since 1978 and another in McIntosh, Alabama, USA [102]. In contrast
to the other pure EES systems, CAES systems have a significant marginal cost dur
ing the discharge phase due to the natural gas consumption. Feasibility of CAES
is constrained by the availability of suitable geology to store air although suitable
sites are speculated to be available in much of North America and Europe [102].
New CAES facilities are proposed to be built in Ohio, Texas, and Iowa [91]. Some
new developments in CAES technology are presented in [103].
3.1.3 Contributions of the Chapter
The objective of the research project presented in this chapter is to develop a model
to support optimal bidding process of a WPP that is in joint operation with a CAES
system and hedges against wind power variability. The model specifically addresses
the uncertainty in the wind resource availability and the market price of electricity
with integrated market risk control. The physical constraints of the wind farm
and the CAES system such as the generation and storage capacity limits, ramp
54
rate constraints, and transmission limitations are also taken into account. While
the main intended application of the model is to support shortterm operations
optimization, the model can also be used to support wind power and CAES invest
ment decisions making. Furthermore, feasibility of energy storage options other
than CAES can also be explored with minimal modifications to the model. With
respect to the related studies in market integration of wind power and EES, the
main contributions of this work are as follows:
• While majority of the studies have assessed the value of EES by taking a long
term generation planning approach, there is a gap in published literature in
studies on shortterm operations optimization of hybrid wind and EES power
generation systems. This chapter contributes to the knowledge in the latter.
• This model differs from the deterministic operations optimization approaches
taken in [101] and the CAES valuation framework presented in [99] because
a stochastic optimization approach is taken, incorporating the uncertainty in
market conditions and wind resource availability.
• The main difference between the approach taken in the model presented in
this chapter and [76] is the integrated risk management technique. This
contribution is significant because the power generation assets are operated
by risk averse, profit seeking agents.
• This work also differs from [76] because the EES system studied in this chap
ter has a significant marginal cost during the discharge phase, which signifi
cantly affects the economics of EES operations.
55
3.2 Operations Optimization Under Uncertainty: Problem Descrip
tion and Solution Approach
In this section the operations optimization problem of the wind power and EES
system and the proposed solution approach are described. To retain the simplicity
of the model and for the clarity of discussion, a hypothetical power market that is
organized as a DAM with a single round of bidding is considered in the preceding
sections of this chapter. The system under study is a WPP and CAES system jointly
participating in the aforementioned DAM (Figure 3.2a). In order to get insights into
the value of adding a CAES system to a WPP, a standalone WPP is also modelled
(Figure 3.2b). In both cases the respective power generation system is assumed to
be a "price taker" so that the market price can be taken as an exogenous parameter
(in other words output of the WPP and the CAES system does not affect the market
price). It is also assumed that hourly bids for the 24 hour period of the following
day are submitted at 12:00. In reality, many power markets have multiple trading
floors and bidding rounds in addition to a DAM [70,73]. The proposed model can
be modified to add further sophistication to fit to conditions of an actual market.
Electric power generators strive to maximize their revenues by setting optimal
production schedules and submitting appropriate bids to power markets. Tech
niques to formulate strategic bidding are an active area of research in power sys
tems and have been well studied (for example see [104–109]). Among other factors
such as production cost, network access constraints, generation unit technical
constraints, and behaviour of other competitive suppliers, a primary factor that
would influence the bid formulation is the market price. Market price will be re
vealed only after market closure, where the accepted and rejected bids are selected.
Therefore, any generator participating in a DAM faces the uncertainty in system
price. In contrast to a conventional dispatchable generator, wind power producers
56
(a) Wind+CAES joint operation (b) Standalone wind
Figure 3.2: Systems under study
12:00 24:00
24:00
day D+1day D
Bidding to (D+1) market
Real time operation
Need recourse against previous decisions
Uncertain wind generation & system price
Figure 3.3: Operations decisions time line
are subjected to uncertainty in both the system price and wind resource avail
ability, which is significant at the time where DAM bids have to be submitted.
Therefore, optimal bidding techniques developed for conventional generators can
not be directly applied to wind generators optimal bidding problem. The strategic
bidding techniques proposed for WPPs, such as the ones in [87,110], can be used
to increase the profits by minimizing imbalance charges. However, their ability to
increase the WPP’s profits are limited to financial hedging and wind power curtail
ment. Use of CAES (or any other EES) provides the WPP operator both physical and
financial flexibility. The operations decision making of a WPP and a CAES system
jointly participate (henceforth referred to as ‘‘wind+CAES system’’) in a DAM can
be modelled as a two stage process as described below.
57
Stage 1: First stage decision is a set of hourly bids to the DAM that would maximize
profits. These bids are to sell either wind/CAES generated energy or to pur
chase energy to store in the CAES system. The operator formulates the bids
by forecasting the wind availability and market price in the period of interest.
At this stage, there is considerable uncertainty in wind power availability and
market price (Figure 3.3). This type of decision making is known as here and
now decisions [111].
Stage 2: Second stage decisions are made after complete or better knowledge (re
duced uncertainty) of wind availability and system price is obtained. For
example, a significantly accurate wind power forecast can be made about an
hour ahead of the actual delivery hour [74]. At this stage, a cleared bid can
not be altered. But depending on the observed realization of random events,
CAES system operations can be rescheduled, providing recourse against stage
1 decision (ie. bids). As the decisions are made after the required information
is obtained, these type of decisions are called wait and see decisions [111].
The salient feature of this decision making process is that profit maximization
has to be done using information with significant uncertainty. Therefore, a suitable
optimal decision modelling method is taking a ‘‘stochastic programming’’ approach.
3.2.1 Stochastic Programming Solution
Stochastic programming (SP) is a branch of optimization where some or all pa
rameters in the objective function or/and constraints are random variables [112].
Knowledge of the distribution of those random variables is needed to solve SP prob
lems [111]. Different SP models and solution methods for SP problems can be found
in [111–113]. Examples of SP applications in electric power systems can be found
in [76,88,114–116].
58
}
1st stage decision
(Day ahead bids)
s
2nd stage decisions
(Wind farm and
CAES operation )
Uncertain events
are revealed
Figure 3.4: Two stage decision making process
The decision making problem encountered in wind+CAES system operations
optimization can be modelled as a type of SP model known as recoursebased
model, where decisions are made in two stages and random events are revealed
in between [111]. This type of model was first introduced in [117]. Development
of a recoursebased SP model for wind+CAES system operations optimization is
described below.
• The objective of the wind+CAES system operator is to maximize the profit
of the wind+CAES system by selling wind/CAES generated electricity in the
DAM. The expenses incurred are payments to electricity purchased from the
market, CAES fuel cost, and any imbalance charges. Similar to [76], the
imbalance charges are assumed to be proportional to the absolute value of
the energy imbalance and settled at market price of the respective hour.
• Uncertainty in wind power and market price is represented by a set of discrete
wind scenarios and price scenarios. For example, a given wind scenario
represents the wind power availability in the following day. As the wind+CAES
59
system is assumed to be a price taker, wind scenarios and price scenarios are
mutually exclusive and are combined to form a single scenario tree (Figure
3.4). A probability of occurrence is assigned to each scenario. These scenarios
are assumed to be generated using wind and price forecasting tools.
• The objective of the model is to maximize the expected profit under each
scenario.
• Outputs of the model are a single set of hourly bids to the DAM (first stage
decision) and a set of WPP and CAES operating rules that provide recourse
against first stage decision at the onset of realization of random events (second
stage decision; Figure 3.4).
• The system operations are constrained by wind and CAES installed capacity
limitations, capacity of the CAES storage cavern, ramping limitations, and
transmission limits. The transmission constrain may be due to a physical
limit or due to a financial transmission contract. A ramp rate limit may be
enforced by the TSO.
• Non fuel variable operating costs are assumed to be negligible.
• The model calculates the energy quantity of energy offered to the DAM in each
hour. As the system is a pricetaker, it is assumed the operator sets the price
to zero or to an appropriate amount ensuring the offered quantity get accepted
in the energy auction. Bid price setting suitable to a pricetaker strategy such
as the one proposed in [104] can easily be adopted for this purpose.
Mathematical formulation of the model is described in section 3.3.
60
← β−VaR = α
β−CVaR →
← mean
Pro
babi
lity
Profit, Bs
Figure 3.5: βVaR and βCVaR of a profit distribution Bs
3.2.2 Risk Management
Risk management is an essential part of power trading as electricity generation
companies are profit seeking entities [118,119]. Because of the inherent uncer
tainty in wind power availability, the wind+CAES system is at a significant risk
of having to settle imbalances. Therefore, it is important to integrate risk man
agement measures into wind+CAES system operations optimization model. Risk
management can be achieved by concurrently optimizing a risk measure along with
maximizing the expected profit [119]. Standard risk measures that are used for
risk management include the standard deviation, mean absolute deviation (MAD),
valueatrisk (VaR), and conditionalvalueatrisk (CVaR) [119,120]. VaR is a com
monly used risk measure, mainly in the finance and insurance industry. With
respect to a profit function, Bs and a confidence level β, by definition VaR is the
lowest amount α such that, with probability β, the profit will exceed α (called as
β − V aR; see figure 3.5). CVaR (or β − V aR) is the conditional expectation of the
profit below that amount α [120]. In contrast to many other risk measures, CVaR is
proven to be a coherent risk measure with superior mathematical properties such
as convexity, monotonicity, and positive homogeneity [120–122]. It is also suitable
for recoursebased SP models [111,113,120]. Theoretical proofs and mathematical
61
developments of CVaR for optimization application are presented in two seminal
papers by Rockafellar & Uryasev [120,123]. CVaR applications for power trading
and scheduling can be found in [88,114,115,118,124–126]. Maximizing the CVaR
of the profit distribution generated by different wind and price scenarios is used as
the risk management method of the proposed model.
3.3 Model Formulation
Under a certain scenario, s ∈ S, the wind+CAES system generates revenues by
selling wind energy, gwst and stored energy in CAES system, gcst, at market price
rate, πst in each time step t ∈ T. Variable costs incurred are the fuel cost of the
CAES generated electricity (= cf · gcst; where, cf is the marginal fuel cost of the
CAES system) and payments for electricity purchased from the market, dst. Hourly
bid, bjt is an offer either to sell or to purchase energy in each time step, t ∈ T.
Imbalance charges are proportional to the product of the market price of the hour
and the absolute value of the energy imbalance with a penalty factor λ. The risk
neutral objective function for maximizing the profit of wind+CAES system is given
by equation (3.1).
maximize∑
s∈S
ρs · Bjs (3.1)
where, Bjs =
∑
t∈T
[πst · (gwst + gcst − dst) · τ − cf · gcst · τ − λ · πst · |(g
wst + gcst − dst) · τ − bjt |]
where, ρs = probability of the scenario s
τ = optimization time step (=1h)
In the next step, CVaR maximization is added to (3.1). As proved in [120], β −
CV aR and β − V aR of a profit function Bjs, with a discrete probability distribution
62
ρs, can be characterized in terms of the function Fβ(s, αj) on S × R (Eqn. (3.2)).
Maximizing CVaR is equivalent to maximizing Fβ(s, αj) over all (Bj
s , αj) ∈ S × R.
Fβ(s, αj) = αj −
1
(1− β)
∑
s∈S
ρs · [αj − Bj
s ]+ (3.2)
where, [αj −Bjs ]
+ = max(0, αj − Bjs)
The objective function (3.1) is modified to incorporate risk management as (3.3).
The two weighting factors ω1 and ω2 are positive numbers of which the values are
set according to the risk preference of the wind+CAES system operator. When
ω1 = 1 and ω2 = 0, the objective function reduces to the risk neutral case of (3.1).
For risk averse operation: ω2 > 0.
maximize ω1(∑
s∈S
ρs · Bjs) + ω2 · Fβ(s, α
j) (3.3)
Objective function (3.3) is maximized with subject to constraints (3.4)(3.14).
The CAES storage cavern energy balance equations (3.4) and the energy balance
constraint (3.5) ensure that sufficient energy is available in the storage cavern. The
upper limit of the cavern, Emax is set by the size of the cavern and the lower limit,
Emin is set by the minimum cavern pressure at which air may be discharged. Wind
energy sold gwsh in each scenario is constrained by the wind energy generation in
the respective scenario Wsh (3.6). Limitations of the generation and compression
capacities of the CAES system are handled by (3.7) and (3.8). They also ensure
that CAES system is not generating and storing electricity simultaneously using
the binary operating mode variables msh and nsh (3.9). Transmission and ramping
limits are represented by (3.11) and (3.12) respectively. Energy bids to the DAM, bjt
are constrained by the installed wind capacity and compressor and generator ca
63
pacities of the CAES system (3.13). The optimization problem can be implemented
as a mixed integer linear program (MILP).
rst = rst−1 + dst · τ − ηc · gcst · τ, ∀s ∈ S, ∀t ∈ T (3.4)
Emin 6 rst 6 Emax, ∀s ∈ S, ∀t ∈ T (3.5)
0 6 gwst 6 Wst, ∀s ∈ S, ∀t ∈ T (3.6)
mst · Pmin ≤ gcst ≤ mst · Pmax, ∀s ∈ S, ∀t ∈ T (3.7)
nst ·Dmin ≤ dst ≤ nst ·Dmax, ∀s ∈ S, ∀t ∈ T (3.8)
0 6 mst + nst ≤ 1, ∀s ∈ S, ∀t ∈ T (3.9)
mst, nst ∈ {0, 1} (3.10)
0 6 |gwst + gcst − dst| 6 Ptx, ∀s ∈ S, ∀t ∈ T (3.11)
0 6 |(gwst + gcst − dst)− (gwst−1 + gcst−1 − dst−1)| 6 Pramp, ∀s ∈ S, ∀t ∈ T (3.12)
−Dmax · τ ≤ bjt ≤ (Pmax + Pwmax) · τ, ∀s ∈ S, ∀t ∈ T (3.13)
rsNt≥ E0 (3.14)
Optimal CAES storage cavern operation planning spans over the entire planning
horizon. Hence, there is a temporal link between each time period in the planning
horizon Nt (Nt is the number of intervals in the time set T). Therefore, wind+CAES
system operations optimization must be concurrently carried out for all Nt time
periods. Furthermore, constraint (3.14) was added to the model to ensure the final
capacity of the storage cavern, rsNtis equal or greater than the initial capacity, E0.
This avoids the potential exaggeration of the value of CAES compared to the stan
dalone wind case. Riskaverse operations optimization model for the standalone
wind power system operation is formulated in Appendix B (Eqns. (B.1)(B.5)).
64
Table 3.1: Model Parameters Used for the Numerical Example
Parameter Value
Installed capacity of WPP, Pwmax 100 MW
CAES maximum generation/compression limit,Pmax, Cmax 50 MW
CAES minimum generation/compression limit ofthe CAES system, Pmin, Cmin 5 MW
CAES storage cavern maximum capacity, Emax 2400 MWh
CAES storage cavern minimum capacity, Emin 240 MWh
CAES fuel cost, Cf 16.8 $/MWha
CAES electricity input/output ratio, ηc 75%
Trasmission limit, Ptx 150 MW
Ramp rate limit, Pramp 3 MW/minute
Penalty factor for energy imbalances, λ 1
Confidence level for risk management, β 95%
a Fuel cost of the CAES system is calculated assuming an operating heat rate of 4.2 GJ/MWh
(HHV basis) [102] and a natural gas price of $4/GJ.
3.4 Case Study
In this section, a numerical experiment performed with the model is described. The
input parameters used for this experiment are listed in Table 3.1.
3.4.1 Wind and Price Scenario Generation
Wind and price scenarios for the case study are generated using historical data
obtained from the Alberta Electric System. Time series data of hourly wind power
generation and market prices in January of 2008 are obtained from the AESO [127].
In order to generate price scenarios, normally distributed zero mean random noise
is added to hourly pool prices of an arbitrarly selected week day. The standard
deviation of the random noise is assumed to be the mean absolute error of a price
forecast made for a winter weekday (Hourly forecasted and actual pool prices are
65
published by the AESO [128]). Noise values are generated using a random number
generator and a set of 32 price scenarios is produced.
In order to generate wind scenarios, a second order autoregressive (AR2) time
series model is fitted to hourly wind power production data of a WPP in Alberta
using the methods described in [54, 55]. The fitted AR2 model is given by (3.15)
and it assumes that the WPP output of a certain hour, Wt, depends on the output of
the previous two hours. The model parameters φ1, φ2, and σw are estimated using
the Matlab® System Identification Toolbox. This AR2 model along with a random
number generator, is used to produce 32 wind scenarios. These scenarios are
linearly scaled to match the output of a 100MW WPP.
Wt = φ1Wt−1 + φ2Wt−2 + ǫt (3.15)
where, ǫt = random normal noise with zero mean and standard deviation σw
Price and wind scenarios generated using the methods explained above are
depicted in Fig. 3.6. The two sets are combined to form a set of 1024 scenarios
(32× 32) and these scenarios are assumed to be equally probable.
3.4.2 Results and Discussion
The optimization problem characterized by equations (3.3)(3.14) (wind+CAES) and
(B.1)(B.5) (standalone wind) are implemented in Matlab®/Tomlab® environment
and solved using CPLEX 12.1® solver. The imbalance penalty factor λ and the
value of weighting factor ω1 are set to unity for all preceding results (ie. λ = 1 &
ω1 = 1). All costs and prices are expressed Canadian dollars (CAD). The objective
of the case study is to develop insights of the value of adding CAES to the WPP, to
investigate how the operator’s risk preference affects the revenues, and to assess
the value of taking the SP approach for wind+CAES system operations optimization.
66
2 4 6 8 10 12 14 16 18 20 22 240
50
100
150
CA
D/M
Wh
Price Scenarios
Hour
2 4 6 8 10 12 14 16 18 20 22 240
20
40
60
80
100
120
MW
Wind Scenarios
Hour
Figure 3.6: System price and wind power scenarios used for the case studyThese have been generated using historical data from Alberta Electric System.
Expected profit, average imbalance charges, 95%VaR, and 95%CVaR of the
standalone wind and wind+CAES system operations as calculated by SP models are
listed in table 3.2. The models are executed multiple times using the same scenario
set depicted in figure 3.6 to produce results for risk neutral operations and risk
averse operations. For the risk averse operation, the value of ω2 is assumed to be
0.5 (in other words optimizing expected profit is two times as important compared
to maximizing CVaR). Distribution of the profit functions in each case are depicted
in figures 3.7a and 3.7b. As can be seen from table 3.2 (and figure 3.7), in risk
neutral operations the expected profit of the wind+CAES system is 31% higher than
that of the standalone wind system. The average imbalance charges incurred for
the wind+CAES system has significantly dropped to 51% of that of standalone wind
system.
67
Table 3.2: Expected profits, imbalance charges, and risk measures(Values of weighting factors for risk neutral operations: ω1 = 1 & ω2 = 0; risk averse
operations: ω1 = 1 & ω2 = 0.5)
Standalonewind operation(CAD)
Wind + CAESoperation(CAD)
Expected profit:
risk neutral 43,714 57,383
risk averse 42,976 56,930
Expected imbalance charge:
risk neutral 5,522 2,711
risk averse 3,848 1,733
95%VaR:
risk neutral 27,000 39,076
risk averse 32,751 45,091
95%CVaR:
risk neutral 25,688 37,976
risk averse 31,312 42,677
68
20 25 30 35 40 45 50 55 60 650
20
40
60
80
100
120
Standalone Wind (Risk neutral, ω2=0)
Profit (thousand CAD)
Fre
quen
cy
Profit distribution
95%−VaR
95%−CVaR
Expected profit
20 25 30 35 40 45 50 55 60 650
20
40
60
80
100
120
Standalone Wind (Risk averse, ω2=0.5)
Profit (thousand CAD)
Fre
quen
cy
(a)
35 40 45 50 55 60 65 70 750
20
40
60
80
100
120
Profit (thousand CAD)
Fre
quen
cy
Wind+CAES (Risk neutral, ω2=0)
Profit distribution
95%−VaR
95%−CVaR
Expected profit
35 40 45 50 55 60 65 70 750
20
40
60
80
100
120
Wind+CAES (Risk averse, ω2=0.5)
Profit (thousand CAD)
Fre
quen
cy
(b)
Figure 3.7: Profit distributions of: (a) standalone wind power plant operation;(b) Wind + CAES joint operation
69
0 5 10 15 20 250
20
40
60
80
100Optimal Generation Schedule (Standalone Wind)
MW
h
Hour
Risk neutral (ω
2=0)
Risk averse (ω2=0.5)
0 5 10 15 20 250
20
40
60
80
100Optimal Generation Schedule (Wind + CAES)
MW
h
Hour
Figure 3.8: Energy bids to the dayahead electricity markets
Energy bids for the DAM of the two configurations (stage 1 decision) are depicted
in figure 3.8. Comparison of the scenario set (Figure 3.6) and the bids (Figure 3.8)
show that for the wind+CAES system strategically arbitrage energy to maximize
profits. As described in section 3.3, the models generate WPP and CAES system
recourse operating rules (stage 2 decision) that can be followed as the delivery hour
approaches and better knowledge of wind and price are obtained. For example,
Figure 3.9 depicts the CAES operation if the actual wind generation and the market
price are revelled to be those of the scenario #48. This figure further illustrates the
wind+CAES system operator’s strategic decision to store wind energy rather than
selling during the periods with low market prices and vice versa.
3.4.2.1 Risk Averse Operation
In the case of riskaverse operation the power plant operator gives up some of the
expected profit to increase the 95%CVaR and consequently 95%VaR. At a same
70
0 5 10 15 20 250
20
40
60
80
100Wind Power Generation and Market Price
Wind power (MWh)
Market price (CAD/MWh)
0 5 10 15 20 25−60
−40
−20
0
20
40
60
MW
h
CAES Operation
Risk neutral
Risk averse
0 5 10 15 20 250
10
20
30
40
50Wind Energy Stored in CAES
MW
h
0 5 10 15 20 250
10
20
30
40
50Grid Energy Stored in CAES
MW
h
Hour
Figure 3.9: Operation of the CAES system under the scenario #48.The first subfigure from the top shows the wind power available and market price. Thesecond subfigure shows the CAES system operation under this scenario. In this figure,positive values represent energy sales to the grid (discharge mode) and negative valuesrepresent energy storing (storage mode). The third and fourth subfigures depict the windenergy stored in CAES and grid energy stored in CAES system (electricity purchased fromthe market) respectively.
71
37 38 39 40 41 42 43 44 4554
55
56
57
58
ω2 = 0 ω
2 = 0.25
ω2 = 0.5
ω2 = 0.75
ω2 = 1
ω2 = 2
β − CVaR (thousand CAD)
Exp
ecte
d P
rofit
(tho
usan
d C
AD
)
Wind + CAES (β = 95%)
25 26 27 28 29 30 31 32 3342
42.5
43
43.5
44
β − CVaR (thousand CAD)
Exp
ecte
d P
rofit
(tho
usan
d C
AD
)
ω2 = 0
ω2 = 0.25
ω2 = 0.5
ω2 = 0.75ω
2 = 1
ω2 = 2
Standalone Wind (β = 95%)
Figure 3.10: Efficient frontiers of the two power plant configurations
level of risk preference (ω2 = 0.5), in both configurations significant improvements
in 95%CVaR and 95%VaR are obtained at loss of a marginal amount of profit.
The observed profit loss compared to the risk neutral operation of the wind+CAES
system is 0.8% and that of the standalone wind system is 1.8%. Consequently the
former has better performance. The other important feature is the magnitude of
95%VaR compared to the expected profit. In the risk averse operation, 95%VaR
of the wind+CAES system is 80% of the expected profit and that of standalone wind
system is 76%. Therefore, the chances of achieving lower profits have been lowered
by adding the CAES system to the WPP.
The trade offs between profit maximizing and operator’s risk preference are
further investigated using efficient frontiers (expected profit vs 95%CVaR curve).
Efficient frontiers are generated by running the models multiple times with different
ω2 values while keeping ω1 at unity. As can be seen from figure 3.10 the expected
72
Table 3.3: Expected value of perfect information (EVPI) for the wind+CAES system(EVPI is expressed as a percentage of the expected profit of the full stochastic solution at
respective risk aversion level)
Perfect knowledge of
wind & price wind only price only
Risk neutral (ω2 = 0) 13% 7.2% 3.2%
Risk averse (ω2 = 0.5) 14% 7.6% 2.4%
profit of the standalone wind case drops more sharply as the system operator
becomes more risk averse. Therefore, adding the CAES system provides more
operational flexibility in terms of risk management. Another observation is that
attempting to increase CVaR beyond a certain point provides diminishing results
and significant profit losses. Hence, the profitCVaR trade off must be appropriately
made, based on an efficient frontier, to avoid unnecessary profit losses.
3.4.2.2 Expected Value of Perfect Information
In this section, the expected value of perfect information (EVPI) of the wind+CAES
system is investigated. With respect to this case study, EVPI measures the maxi
mum amount the power plant operator would be ready to pay in return for complete
(and accurate) information about future wind power availability, market price, or
both. In a recoursebased SP problem, the EVPI is the difference between the ex
pected value of ‘‘wait and see’’ solutions (EWS) and the stochastic solution [112]. In
order to calculate EWS first the profit produced by each of 1024 scenarios (perfect
knowledge of both price and wind) is calculated by running separate deterministic
optimization problems and then taking the mean of those profit values. EVPI of
perfect knowledge of wind, price, and both are listed in table 3.3. These values
are expressed as a percentage of expected profit calculated by the SP model in
respective cases. Some observations merit further analysis. As expected, a risk
73
averse operator is willing to pay more for perfect knowledge of future events. The
value of knowledge of wind availability is more important than knowledge of mar
ket price. This is because the marginal cost of wind energy is zero and arbitraging
wind energy is more profitable than electricity purchased from the grid. Therefore,
foreknowledge of wind power availability can be used for better arbitrage planning.
Furthermore, since wind energy is not controllable (the only option being curtail
ment), the impact of wind power uncertainty on energy imbalances is higher. The
other interesting observation is the value of perfect knowledge of price for the risk
averse operator is marginally lower than that for the risk neutral operator. The
reason for that is the fact that the strategy taken by the risk averse operator is to
bid more conservatively irrespective of the market price, and that may reduce the
value of foreknowledge of price. Two general conclusions can be drawn from these
results. Significant EVPI values (close to 10% of expected profit) mean that the
uncertainty of available information plays an important role in the decision making
process. This affirms the value of taking a stochastic decision modelling approach.
A second conclusion is the fact that even with access to a physical hedging tool,
with respect to DAM participation, it is still important to have an accurate wind
forecast. The ability to participate in intraday balancing market may change that
conclusion, and needs further investigation.
3.4.2.3 Sensitivity of CAES Parameters
A sensitivity analysis was carried out to study the influence of CAES parameters.
CAES generation/compression capacity, storage cavern size, CAES efficiency and
natural gas price were independently varied from the values listed in table 3.1.
The number of scenarios used is reduced to 400 (20 wind/ 20 price) to lower the
simulation time. Efficient frontiers corresponding to each parameter value are
depicted in figures 3.11a3.11d. As can be seen from them, wind+CAES system
74
36 38 40 42 44 46 48 50 52 5450
51
52
53
54
55
56
57
58
59
60
β−CVaR (thousand CAD)
Exp
ecte
d pr
ofit
(tho
usan
d C
AD
)
Storage hours = 6hStorage hours = 12hStorage hours = 24hStorage hours = 48h
(a)
36 38 40 42 44 46 48 50 52 5450
52
54
56
58
60
62
64
66
β−CVaR (thousand CAD)
Exp
ecte
d pr
ofit
(tho
usan
d C
AD
)
Pmax
/Cmax
= 25MW
Pmax
/Cmax
= 50MW
Pmax
/Cmax
= 100MW
(b)
36 38 40 42 44 46 48 50 52 5450
51
52
53
54
55
56
57
58
59
60
β−CVaR (thousand CAD)
Exp
ecte
d pr
ofit
(tho
usan
d C
AD
)
Electricity input/output ratio = 0.85Electricity input/output ratio = 0.75Electricity input/output ratio = 0.65
(c)
36 38 40 42 44 46 48 50 52 5450
51
52
53
54
55
56
57
58
59
60
β−CVaR (thousand CAD)
Exp
ecte
d pr
ofit
(tho
usan
d C
AD
)
NG price = $2/GJNG price = $4/GJNG price = $6/GJNG price = $8/GJ
(d)
Figure 3.11: Influence of CAES parameters on Wind+CAES system economicsParameters varied are as follows: (a) storage cavern size (in storage hours at installedcapacity); (b) generation/expansion capacity; (c) CAES electricity input/output ratio; (d)natural gas price. All CAES parameters, except the one varied, are kept at set to the samevalues listed in table 3.1.
economics are most sensitive to CAES generation/expansion capacity and natural
gas price. The least sensitive parameter is the CAES storage cavern size (expressed
in storage hours at 100MW generation/expansion capacity). As can be seen from
figure 3.11b, the expected profit gain is marginal beyond storage time of 24 hours
and there is no gain after 48 hours (curves for cavern sizes beyond that are not
shown). This result is an artifact of the planning period used and the constraint
requiring the final capacity of cavern to be equal to or greater than the initial
capacity. Since a planning period of 24 hours is used, the optimization model does
not see any value in storing energy beyond that period.
75
3.5 Conclusions
The variability and uncertainty of the output may put a WPP participating in con
ventional electricity markets at risk of making lower profits due to a low correlation
between wind power production and market price and having to settle energy im
balances. Integrating a CAES system with a WPP is an effective solution to mitigate
these challenges. This chapter proposed a two stage stochastic programming model
that can be used to assist optimal operations decision making and managing fi
nancial risks of a WPP and a CAES system jointly participate in a DAM. A case
study with realistic wind power and market price conditions is also presented to
demonstrate the feasibility of the model to assist operational decision making. The
stochastic optimization approach is found to be appropriate for the decision mod
elling situation studied in this chapter.
The CAES system provides the WPP more operational flexibility and a means to
increase the economic value of wind power. An interesting and necessary future
study is to assess whether the increase in profits of the joint operation is sufficient
to make the additional investment for the CAES system. The model presented,
along with appropriate wind power and price scenarios can be used to do such a
detailed investment assessment under the uncertainty of future electricity market
prices, wind resources, and fuel price. That type of analysis can be used to op
timally size installed wind capacities, CAES system capacities, and transmission
capacities.
Some directions for further investigations and future enhancements of the
model are stated below.
• Current implementation determines only the quantity of energy bids. Price
determination capability can be easily implemented, for example by using the
method proposed in [104].
76
• Further analysis should be made to determine optimal planning horizon and
to provide insights of the required model constraint adjustments.
• A more comprehensive implementation of the model should include the oper
ations planning capability to participate in intraday balancing markets and
ancillary service markets.
77
Chapter 4
Evaluating the Role of Cogeneration for Carbon
Management in Alberta
4.1 Introduction
The various carbon emissions management policies being discussed or adopted
around the world create a unique set of experiments in policy, engineering and
economic pricing. All else being equal, an economically efficient policy should
create a single economy wide marginal carbon price signal either in direct form,
such as a carbon tax, or in an implied form such as a cap and trade system.
In either case the objective is to influence energy sector investment and decision
making so as to costeffectively restrain emissions. Of course, restraining emissions
is but one objective of government policy; and, there may be sensible reasons to
deviate from economywide approaches. If, for example, there is reason to believe
that imposing a relatively high carbon price will spur technical innovation in a
particular sector lowering the future cost of emissions abatement so substantially
as to make up for the shortterm loss of economic efficiency.
Theory aside, in most cases policy makers have opted to use complex facility or
productbased policy tools that reflect political pressure against enacting efficient
economywide carbon policies. Enforcement of such policies require emissions
accounting methods that are data and management intensive. Furthermore, choice
of facility or productbased carbon accounting methods is inherently arbitrary in
the sense that there are no simple general rules for producing emissions estimates
which (a) produce stable results and (b) are selfconsistent in the sense that the
78
total emissions from a set of facilities are independent of the way the rules are
applied. This arbitrariness can be an impediment to academic assessment of life
cycle emissions, but when such emissions calculations are used as part of policy
then one can expect rational profitseeking firms to exploit the arbitrariness to
reduce their burden under the emissions control policy.
In this chapter we examine emissions rules for oil sands producers in the Cana
dian province of Alberta, as an example of a case where uncertainty in emissions
accounting and the burden of administrative complexities has interacted to frus
trate efficient carbon policy. These concerns are particularly relevant for a facility
with multiproduct outputs, such as a cogeneration facility that produces both
electricity and steam for bitumen production.
Oil sands operations in Alberta are playing an increasingly important role in
North American oil supplies and Canada’s oil export market. Production of bitu
men, the primary hydrocarbon extracted from oil sands, reached approximately
1.3 million barrels per day in Alberta in 2008, satisfying approximately 1.6% of
world demand of oil [129,130]. Bitumen recovery and processing requires a sig
nificant amount of thermal energy and electricity [130]. Natural gas is the main
fuel currently used to satisfy the thermal energy demand of oil sands operations.
In 2003, the volume of natural gas purchased from Alberta’s gas market for the
purposes of bitumen recovery and upgrading amounted to 5.2 billion cubic metres,
roughly 5% of Canadian demand and 14% of demand in Alberta [130]. The high
energy intensity of oil sands operations combined with the fact that the primary
energy sources used to generate heat and electricity are predominantly fossil fuels,
results in relatively high greenhouse gas (GHG) emissions from this sector. It has
been reported that the oil sands sector contributed approximately 5% of Canada’s
emissions resulting in 37.2 million tCO2 equivalent (CO2 eq.) in 2008. This is a
79
39% growth from the oil sand sector’s GHG emissions in 2000 [45].
Cogeneration, the combined generation of electric power and thermal energy,
provides an option for oil sands operations to meet both steam and electric energy
demands onsite. Though various configurations are possible, oil sands operations
typically use a gas turbine to generate power coupled with a heat recovery steam
generator (HRSG) that captures waste heat from the gas turbine exhaust to produce
steam or hot water [131]. Despite higher onsite fuel use, cogeneration has a high
operating efficiency, on the order of 7080%, compared to standalone steam and
electricity production. The primary requirement to justify the incorporation of a
cogeneration system is the presence of a steady thermal energy demand. Due to
the substantial heat requirements in oil sands operations, electricity production of
a cogeneration system incorporated into an oil sands operation typically exceeds
the onsite demand, which may result in electricity exports to the Alberta grid.
Alberta’s electricity sector, where the generation is dominated by coal and natural
gas, produced 52 million tCO2 in 2008 making it the most carbon intensive power
system in Canada [45]. In 2008 the combined GHG emissions of Alberta’s oil sands
sector and the electricity sector amounted to 37% of the province’s 244 million
tCO2 eq. emissions. The growing oil sands sector has the potential to increase its
cogeneration capacity, potentially displacing higher carbon intensive electricity in
the electricity sector of Alberta.
In this chapter we examine the use of cogeneration for oil sands operations
in the context of carbon emissions management. Our main objectives are to: (1)
assess the role of cogeneration for carbon emissions reduction in Alberta; (2) in
vestigate the effect of present GHG emissions reduction regulation in Alberta on
the economics of cogeneration; (3) evaluate the efficiency of current and alterna
tive emissions control policies; and, (4) examine the way in which uncertainties of
80
facility or productbased carbon accounting complicates efficient carbon policy.
4.2 Background
4.2.1 Oil sands operations
The proven oil sands reserves in Alberta are estimated at 170 billion barrels of
crude bitumen. In 2006, Alberta’s oil sands were the source of about 62% of the
province’s total crude oil (and equivalent) production and about 47% of all crude
oil (and equivalent) produced in Canada. Forecasts of bitumen production growth
leads to a production level as high as 3 million barrels per day by 2020 and up to
5 million barrels per day by 2030 [130,132]
Table 4.1: Electricity and natural gas demand for bitumen extraction and upgrading.
ProcessNatural Gas
(GJ/bblbitumen)
Electricity(kWh/bblbitumen)
Extraction:Mining 0.30.4 1416Insitu 11.6 115
Upgrading 0.150.45 1455
Source:[131,133]
Oil sands operations consist of extracting bitumen and in some cases upgrading
that into synthetic crude oil. Both phases need a substantial amount of energy, the
amount of which depends on extraction technology, among other things. Currently,
the principal extraction technologies in use can be categorized as surface mining
and in situ extraction techniques [132]. The former removes the oil sands by
mining and extracts the bitumen through a series of processes utilizing thermal
energy and water. The latter involves drilling wells and injecting steam to reduce
the viscosity of bitumen so it can be pumped to the surface. The two main thermal
81
insitu techniques that are in commercial use are ‘‘cyclic steam stimulation (CSS)"
and ‘‘steam assisted gravity drainage (SAGD)". Short to medium term bitumen
production growth is forecasted to occur mainly using mining and SAGD extraction
technologies [130]. The energy demands for bitumen extraction and upgrading are
listed in Table 1.
A reliable supply of electricity and thermal energy is critical for both bitumen
extraction technologies. Currently, all mining and upgrading projects that are in
commercial operation have incorporated cogeneration while only 6 out of 25 com
mercially operating insitu extraction projects (including both SAGD and CSS) have
installed cogeneration systems. However, those 6 projects represent approximately
65% of the total in situ bitumen extraction [134]. The installed cogeneration ca
pacity in mining and upgrading operations amounted to 1446MW in 2010 that
generated 9076GWh of electricity of which 77% was consumed onsite. Thermal in
situ production had 908MW of installed capacity and generated 6615GWh in 2010,
of which 51% was consumed onsite [130].
According to a recent survey, the factors that are critical in an oil sands oper
ators’ decision to invest in cogeneration include capital costs, the price of natural
gas and electricity, security and reliability of electricity supply, environmental per
formance of the operation, present and future GHG control regulations, and cost
and availability of transmission [135]. The same survey reports a tendency to delay
the cogeneration investment and also size capacity sufficiently to satisfy only the
host facilities electricity demand in light of uncertainty associated with the factors
listed above.
4.2.2 Alberta electric power system
At the end of 2010, Alberta’s electric power system had 13,071 MW of installed
generation capacity, which produced 70,586 GWh of electricity. Coalfired elec
82
tricity, currently supplying primarily baseload generation, represented 44% of the
installed capacity and 58% of total generation in 2010. Natural gas fired electric
ity (from simple cycle, combined cycle and cogeneration technologies) represented
40% of installed capacity and 34% of total generation in 2010 [130]. Approximately
75% of the installed natural gas fired generation capacity is cogeneration. The ma
jority of the remaining installed generation capacity consists of renewable genera
tion technologies, including wind, hydro and biomass. The ‘‘deregulated" Alberta
power system has opened up the generation and retail electricity sales for competi
tion while the transmission system remains regulated. The competitive generation
market environment allows cogeneration system operators to sell excess electricity
in the Alberta’s wholesale electricity market. The transmission links that connect
the oil sands regions to the rest of the Alberta grid currently have a maximum
import/export capacity of 600MW. The Alberta Electric Power Systems Operator
(AESO), however, is planning to expand the transmission capacity serving the oil
sands region within next 56 years [46].
Since electricity generation in Alberta is dominated by fossil fuels, particularly
coal, the average grid electricity has a very high carbon intensity ( approximately
0.84 tCO2/MWh) compared to the other Canadian provinces. Electricity generation
in the province produced 52 million tonnes of CO2e in 2008 which is approximately
21% of Alberta emissions1, the largest contribution from a single economic sector
in the province [45]. The magnitude of emissions, cost of emissions control, and
the efficiency of regulation with central and limited ownership make the electric
power sector a prime target of GHG emissions reduction targets in Alberta.
The coal generation fraction of the generation base is changing, in part due to
natural attrition from planned retirements. Approximately 1100 MW of coal fired
generation capacity is expected to retire between 2010 and 2020 [136]. Retirement
1This is approximately 7% of total Canadian emissions
83
of these units, along with 23% forecasted demand growth implies a need for new
generation capacity. Thirtythree billion tonnes of discovered coal reserves remain
in Alberta, implying that coal could provide a significant source of electricity for
many years to come [130]. However, a stringent carbon control regulation may
render conventional coal fired generation uneconomic.
4.2.3 Current carbon management policies in Alberta
The province has set goals to reduce the provincial CO2 emissions relative to a
growing baseline by 50 million tonnes by 2020 and by 200 million tonnes by 2050.
The 2050 reduction target represents a 50% reduction below the business as usual
level and 14% below 2005 level [137].
In 2007 the Alberta provincial legislature enacted the ‘‘Specified Gas Emitters
Regulation (SGER)" to regulate GHG emissions. This regulation uses an intensity
and productbased approach. SGER requires facilities in Alberta that have direct
annual GHG emissions larger than 100,000 tonnes CO2e to reduce their emissions
intensity by 12% of facility’s ‘‘baseline emissions intensity (BEI)" [138]. Under
SGER, the emissions intensity is defined as the GHG emissions per unit economic
output of the facility2. Facilities that are regulated by SGER can comply by mak
ing improvements to their operations; by purchasing Alberta based ‘‘offset credits";
by using or purchasing ‘‘emissions performance credits (EPC)"; by contributing to
the ‘‘Climate Change and Emissions Management Fund (CCEMF)" at the rate of
C$15/tCO2e. Facilities that have reduced their emissions intensity by more than
the mandatory 12% reduction target are said to have generated EPCs and these
credits can be banked for future use or be sold to other facilities. The CCEMF is to
be used for projects and new technologies aimed at reducing GHG emissions that
2 For example, for a crude oil production facility, GHG emissions intensity is the total GHG emissions
per one barrel (or 1 m3) of oil produced
84
originate in Alberta. It should be noted that the SGER implicitly caps the price of
carbon in the province at C$15/tCO2e by allowing compliance through contribu
tions to CCEMF at that rate. The SGER has special provisions for facilities with
cogeneration; such facilities are only required to reduce emissions associated with
thermal energy production and the emissions attributed to electricity are exempted
from SGER compliance target. To calculate this, first the BEI for the facility is
set based on the thermal load average over the baseline time period, and then
reference baseline emissions are derived by assuming heat was supplied by a hy
pothetical 80% efficient boiler3. The ‘‘net emissions intensity (NEI)" of the facility, in
a year where the facility has to comply with SGER, is calculated considering only
the emissions associated with thermal energy by subtracting an amount called
‘‘deemed emissions attributed to electricity" from the total emissions associated
with onsite energy production. Deemed emissions attributed to electricity is calcu
lated by multiplying the amount of onsite cogenerated electricity by the emissions
intensity of a natural gas fired CCGT unit, which the SGER guidelines considers to
be 0.418 tCO2e /MWh [138,139]4.
4.3 Model Description
In order to assess the potential for CO2 emissions reductions of cogeneration and
the effects of different GHG emissions management policies on the economics of
cogeneration, we develop a model based on mass and energy balances of two op
tions that satisfy the steam and electricity demands of a SAGD bitumen extraction
3The SGER guidelines do not specify whether this is based on a lower or higher heating value
(HHV). In our analysis we assumed the baseline boiler efficiency to be 80% in HHV.4 Facilities with cogeneration are classified as ‘‘standalone facilities" and ‘‘integrated facilities" and
different guidelines are set under SGER to calculate the emissions intensities. A facility is considered
as a standalone facility if the cogeneration system is the only thermal energy source of the facility
and a facility with other thermal energy sources in addition to the cogeneration system is considered
as an integrated facility. See [139] for full details.
85
Boiler
B
Fuel, FB
Steam, H
Grid Electricity, E
Oil sands operations
Alberta Grid
Feed water, Hfw
(a)
Gas Turbine
T
Heat Recovery Steam
Generator (HRSG)
Fuel, FT Exhaust
Fuel, FGSteam, H1
G
R
Alberta
Grid
Oil sands
operationsEexp Electricity, Ec E
Supplementary Boiler
SB
Fuel, FSB
Steam, H2
Feed water, Hfw1
Feed water, Hfw2
(b)
Figure 4.1: (a) Baseline option and (b) cogeneration option
operation with a production capacity of 30,000 bbl/day. SAGD extraction is used
for this illustrative example for two reasons. First, the steam demand of insitu
extraction methods such as SAGD is higher than mining extraction while the elec
tricity demand is lower. Due to the need for a continuous steam supply and the
moderate electricity demand, insitu extraction plants have a higher potential to
use cogeneration and export electricity to the grid. Second, about 80% of the es
tablished crude bitumen reserves are considered to be buried too deep to mine,
thus we assume that insitu techniques will be used to extract a larger fraction of
the reserves. Of all commercially proven insitu extraction techniques, presently
SAGD has the highest growth rate [130].
In the first option, electricity demand is satisfied through grid electricity im
ports, and steam demand is satisfied through an onsite natural gas fired boiler
86
Table 4.2: Parameters used for the energy and CO2 emissions calculations.
Parameter Value
Bitumen production capacity 30,000 bbl/daySteam demand of bitumen extraction 1.3 GJ/bblElectricity demand of bitumen extraction 12 kWh/bblElectricity production capacity (cogeneration system) 85 MWeMaximum steam production capacity:
Baseline option boiler 1600 GJ/hSupplementary boiler 500 GJ/hHRSG (cogeneration system) 1200 GJ/h
Energy conversion efficiencies (HHV basis)a:Boiler / Supplementary boiler, ηB 85%Gas turbine electricity generation, ηT 30%HRSG heat recovery, ηR 50%HRSG supplemental firing, ηG 95%
Fuel carbon intensities (HHV basis):Natural gas, Icng 0.05 tCO2/GJCoal, Iccoal 0.1 tCO2/GJ
a A sensitivity analysis was done to investigate the effect of the variations ofconversion efficiencies. Through this analysis we found that our conclusionsremain unchanged within the reported range of conversion efficiencies.
with an 85% higher heating value efficiency (henceforth referred to as baseline
option; see Figure 4.1a). In the second option a cogeneration system is used to
produce both electricity and steam (henceforth referred to as cogeneration option;
see Figure 4.1b). It is assumed that the cogeneration system produces excess elec
tricity, which will be sold to the grid and onsite steam demand is satisfied through a
combination of the cogeneration system and a supplementary boiler. The cogener
ation system consists of a gas turbine and a heat recovery steam generator (HRSG).
The HRSG has supplemental firing (also known as duct firing); it can directly fire
fuel in addition to recovering heat from gas turbine exhaust to produce steam [140].
The fuel used in both the baseline option and the cogeneration option is natural
gas. The parameters assumed for the model are listed in Table 4.2. Parameters
87
specific to the boilers and the cogeneration system were obtained from the speci
fications and the test results published by the manufacturers [140]. Capacities of
boilers and cogeneration system were selected to be representative of the typical
sizes and conditions that are in use in oil sands operations [141]. In order to per
form this analysis, we assume that sufficient transmission access is available to
export cogenerated electricity to the Alberta electric system. The transmission sys
tem expansion plan of the AESO supports this assumption [46]. We also assume
that the cogeneration system produces electricity and steam at rated capacity. The
supplementary boiler is used to meet the steam demand not satisfied by the cogen
eration system. The bitumen extraction plant is assumed to be in operation 90% of
the time of a given year. The fuel demands of the baseline option (figure 4.1a and
the cogeneration option (figure 4.1b) are calculated using equations (4.1) (4.5)).
88
FB =H −Hfw
ηB(4.1)
FT =3.6Ec
ηT(4.2)
FG =H1 −Hfw1 − (1− ηT ) · FT · ηR
ηG(4.3)
Fcogen = FT + FG (4.4)
FSB =H −H1 −Hfw2
ηB(4.5)
Where, FB = fuel input to the baseline boiler (GJ/h);
FT = fuel input to the gas turbine (GJ/h);
FG = fuel input to the HRSG (GJ/h);
FSB = fuel input to the supplementary boiler (GJ/h);
EC = electricity produced by the cogeneration system (MWh/h);
H = enthalpy of the steam produced by baseline boiler (GJ/h);
H1 = steam produced by cogeneration system (GJ/h);
H2 = steam produced by auxiliary boiler (GJ/h);
Hfw = baseline boiler feed water enthalpy (GJ/h);
Hfw1, Hfw2 = HRSG/supplementary boiler feed water enthalpy (GJ/h);
ηB = baseline/supplementary boiler efficiency;
ηT = electricity generation efficiency of the gas turbine;
ηG = HRSG supplemental firing efficiency;
ηR = HRSG heat recovery efficiency.
In this analysis, we only consider the CO2 emissions from direct fuel combus
tion for steam and electricity production. Upstream life cycle emissions and the
other GHG emissions are excluded from the analysis. The CO2 emissions from
steam production in the baseline option are calculated by multiplying FB by the
CO2 intensity of natural gas (Icng), assuming complete fuel combustion. The same
method is used to calculate the CO2 emissions associated with the supplemen
tary boiler of the cogeneration option. Estimating total CO2 emissions of the co
generation system is straightforward. However, determining the CO2 emissions
89
associated with electricity alone is not a straightforward calculation as the cogen
eration system produces two energy products with a single stream of input fuel. In
the realm of life cycle assessment (LCA) studies, this accounting complexity that
arise in case of processes with multiple inputs and/or outputs is known as the
‘‘allocation problem" [142, 143]. The theoretical details and guidelines to address
the allocation problem, including structured approaches to choose a method to
allocate process inputs among outputs, are well studied and published, for ex
ample [142,144–151]. However, the fact that there are many methods to address
the allocation problem has led to continued debate among LCA practitioners on the
choice of allocation method [146,152]. We adhere to the common finding that there
is no one best method and consequently, explore the implications of four allocation
methods for the cogeneration case, henceforth referred to as M1, M2, M3 and M4.
This approach is know as ‘‘allocation by physical causal or other relationship" to
solve the allocation problem [142, 143]. The fuel chargeable to electricity (FCE;
in GJ/MWh representing the amount of fuel allocated to electricity) under each
allocation method is calculated using equations (4.6) (4.9).
Method M1 (Eqn. (4.6)) is based on the additional fuel consumed in the co
generation case to produce electricity compared to the baseline option. Under
this method, fuel that would have been consumed by the boiler in the baseline
option—the most likely method to produce steam if a cogeneration system was not
employed—to produce an amount of steam equivalent to the HRSG output (ie. H1) is
allocated to steam. The difference between the total fuel consumed by the cogener
ation system and the fuel allocated to steam is assigned to cogenerated electricity.
This method is also known as ‘‘displacement allocation" in LCA literature [143,144].
Under the M2 method fuel is allocated in proportion to the amount of energy
contained in the two useful products (steam and electricity) of the cogeneration
90
system (Eqn. (4.7)). This ‘‘energy allocation" method is simple and straightforward,
but focuses only on the quantity of energy, ignoring the fact that electrical energy
is higher in quality than steam.
The M3 method takes both the quantity and the quality of the two energy prod
ucts by allocating fuel in proportion to exergy in each product (Eqn. (4.8)). Exergy
of the steam produced is calculated by multiplying the steam enthalpy by the ex
ergetic temperature factor, τ [150]. Since exergy of steam depends on the steam
temperature (T) and the reference environment temperature (T0), FCEM3 is linked
to the operating conditions.
The M4 method allocates fuel in proportion to the economic value of the prod
ucts (Eqn. (4.9)). In this analysis, the economic value of electricity (pe) is set to
be equal to the average price of electricity, which is assumed to be $50/MWh. The
economic value of steam (ph) is assumed to be the average cost of 1GJ of steam
produced by the baseline boiler at natural gas price of $5/GJ (in this case ph =
$4.30/GJ). The CO2 emissions intensity of cogenerated electricity (Icogen) under a
given allocation method is calculated by multiplying FCE by Icng (Eqn. (4.10)).
FCEM1 =Fcogen − (H1−Hfw1)/ηB
Ec
(4.6)
FCEM2 =
(
Ec
Ec +H1
)
· Fcogen ·1
Ec
(4.7)
FCEM3 =
(
Ec
Ec + τ ·H1
)
· Fcogen ·1
Ec
(4.8)
where, τ = 1− T0/T
FCEM4 =
(
pe · Ec
pe · Ec + ph ·H1
)
· Fcogen ·1
Ec
(4.9)
Icogen = FCEMx · Icng; where x = 1, 2, 3, 4 (4.10)
Woffset = (Ioffset − Icogen) · Ec · u · 8760 (4.11)
91
The CO2 emissions offset is calculated using equation (4.11). Here we assume
that cogenerated electricity displaces more carbon intensive electricity in the Al
berta electric power system. The offset amount is determined by Icogen, and the
CO2 emissions intensity of displaced electricity, Ioffset. In a ‘‘deregulated" electric
ity market such as in Alberta, determining which electricity generators are being
displaced by cogeneration units with a high degree of certainty is not possible, as
generation dispatch information is kept confidential. Thus we provide reasonable
estimates that can be made using publicly available data. We investigate the im
plications of four electricity displacement scenarios referred to as S1, S2, S3, and
S4.
Scenario S1 assumes Ioffset to be the average CO2 emissions intensity of the
Alberta electric system. Average CO2 intensity of the Alberta electric system for the
period 20002008 was calculated using the data published by the AESO [153] and
the calculation details are presented in Appendix D.
Scenario S2 assumes that cogenerated electricity, when dispatched, displaces
the units operating at the margin of the generator dispatch stack. In a competitive
electricity market environment, the system operator dispatches different genera
tors to meet the demand following a cost minimization that takes in bids from
participating units. The bid price of the last unit dispatched becomes the system
price of that particular hour, thus called the price setting unit. We assume that
for every MWh of cogenerated electricity, another MWh is backed off from the unit
operating at the margin. The CO2 emissions intensity of the operating margin for
the period from 2000 to 2008 is calculated using the price setting data published
by the AESO [153] 5.
5 In its ‘‘2008 Annual Report" the AESO reports the percentage of the time a certain fuel or generation
technology (coal, natural gas, hydro etc.) set the system price and we assume that the particular fuel
or technology operated in the margin for the same amount of time. However, the data are aggregated
92
The third scenario, S3, assumes that cogenerated electricity displaces coal fired
base load units. As the cogeneration units follow the thermal load of the host fa
cility, they may very well operate as base load generators, bidding appropriately
during peak load and offpeak load hours. Hence it is plausible that they may
displace coal fired units. Scenario S4, following the SGER, assumes that cogen
erated electricity displaces natural gas fired combined cycle gas turbine (CCGT)
generators.
In order to determine the cost of CO2 mitigation from cogeneration and also
to investigate how the cogeneration system economics are affected by CO2 man
agement policies, an engineering economic analysis is developed. We include only
the capital and operating costs to procure energy for bitumen extraction assuming
that project development (drilling, land lease etc.) and non energy related operat
ing costs are identical for both baseline option and cogeneration option. The main
cost parameters assumed for the analysis are listed in Table 4.3. A pre tax 12%
discounting rate was used for the engineering economic analysis. This discounting
rate over a project life of 20 years corresponds to an annual capital charge factor
of 13.3%.
4.4 Results & Discussion
Using the mass and energy balance model we compute the fuel consumption and
CO2 emissions of the two options to satisfy the energy demands of the bitumen
extraction project. Results of the engineering economic analysis and an examina
tion of historic electricity and natural gas prices in Alberta were used to assess the
economic competitiveness of the cogeneration option.
and do not specify which unit is setting the price due to the proprietary nature of such information.
This leads to uncertainties in the calculated emissions intensity as we used a single representative
heat rate value for a given generation technology (see Appendix D for more details)
93
Baseline option Cogeneration option0
0.2
0.4
0.6
0.8
1
1.2
Tot
al s
ocie
ty C
O2 e
mis
sion
s, (
MtC
O2/y
ear)
SAGD project on−site emissionsElectricty sector emissions
Produce steam
Produceelectricity
Producesteam +
electricity
Avoided emissions(31% of the
baselineoption emissions)
SAGD on−siteelectricty demand
Alberta referenceelectricity demand
Figure 4.2: Total CO2 emissions within Alberta, under the two energy optionsThe total CO2 emissions in Alberta, to deliver 124,400 TJ (H) of steam and 650 GWh(Ec) of electricity annually under the two energy options, are presented in thisfigure. The two columns depict the CO2 emissions associated with an identicalamount of steam and electricity. Therefore CO2 emissions from generating anequivalent amount of electricity as in the case of cogeneration option (includingboth electricity consumed onsite and exported to the grid) in the Alberta electricitysystem (which has an average CO2 intensity of 0.84 tCO2/MWh) are added to thebaseline option. No electricity sector emissions are added to the cogeneration optionassuming that cogenerated electricity displaces equivalent amount of high carbonintensive electricity in the Alberta grid. As indicated in the figure, the total Albertaemissions of the cogeneration option are 31% lower than that of the baseline option.
94
Table 4.3: Cost parameters used for engineering economic analysis (all costs arein 2008 Canadian dollars).
Cost parameter Value
Capital costBoiler 400 $/(GJh/h)Cogeneration 1400 $/kWe
Fixed O&M costBoiler 4 $/(GJh/h)Cogeneration 14 $/kWeyear
Variable O & M costBoiler 2 $/GJh
Cogeneration 2 $/MWhe
Natural gas price 210 $/GJElectricity price 0100 $/MWh
The onsite CO2 emissions of the cogeneration option are 42% higher than the
baseline option due to the additional fuel consumed to produce electricity. How
ever, as shown in Figure 4.2, when the CO2 emissions from producing electricity
in the Alberta electric system (an equivalent amount to the electricity generated in
the cogeneration option at an assumed average CO2 intensity of 0.84 tCO2/MWh),
are added to the baseline option to estimate the total emissions, the net CO2 emis
sions of the cogeneration option are 31% lower. However, there is considerable
uncertainty in determining which electricity generating units are being displaced
by cogenerated electricity. Depending on the emissions intensity of the units as
sumed to be displaced, the total Provincial emissions of the cogeneration option
are estimated to be from 6% to 38% lower than that of the baseline option. This is
explored further in section 4.4.2.
4.4.1 CO2 Emissions
The CO2 emissions intensities of cogenerated electricity under different allocation
methods are compared to those of other fossil fuel based electricity, Alberta’s grid
average, and marginal electricity production in Figure 4.3. These results show
95
ABGrid Av. ABGrid OM Coal CCGT M1 (inc. fuel) M2(energy) M3(exergy) M4(economic)0
0.2
0.4
0.6
0.8
1
1.2
tCO
2/MW
h
Grid intensities CogenerationOther fossil fuel basedgeneration
Figure 4.3: CO2 emissions intensities of electricityThis figure depicts the CO2 emissions intensities of the Alberta electric system(average and marginal intensities), coal fired generation, natural gas fired combinedcycle generation, and cogeneration under different allocation methods (M1M4).The grid average intensity calculation considers the energy traded in the Albertaelectricity market and excludes the onsite generation that serves behindthefenceloads (but include the electricity exported to the Alberta grid by behindthefencegenerators such as cogeneration units).
96
that the carbon intensity of cogenerated electricity calculated using any of the
four allocation methods considered is less than the fossil fuel based electricity
generation technologies and the two Alberta grid emission intensities (with the
exception of the cogenerated electricity under the M4 method compared to the
intensity of CCGT).
The choice of allocation method is an important regulatory decision in control
ling emissions from multiproduct output facilities through facility based or product
based regulations. As mentioned in section 4.3, there are many alternative meth
ods to allocate emissions among multiple outputs derived from a common stream
of energy and resources and most of those methods can be rationalized with sound
technical or logical arguments. The allocation method should be chosen consider
ing the context in which allocation is carried out [151]. In case of emissions control,
the regulatory choice of the allocation method should reflect the way the output
products are valued in rational and profit seeking corporate investment decision
making. Therefore, an argument can be made that the allocation method based on
the economic value (M4) should be used where an allocation method is needed for
emissions control regulations. The calculation procedure under M4 method should
consider both the capital cost and operating cost allocations as well as the expected
revenue form the products. This procedure is information intensive and depends on
exogenous parameters. For example in our cogeneration example system, the FCE
under M4 method varies with natural gas and electricity prices. The M1 method
depend on the operating efficiencies of the cogeneration system and also represent
the marginal fuel cost of cogenerated electricity. Hence it can be considered as a
close approximate to economic and technical decision making. Of the four alloca
tion methods investigated, only the M2 method is deemed inferior due to its flaws
discussed in section 4.36. In the remainder of the analysis, where we have to use
6This is not a general conclusion. There can be allocation situations where ‘‘energy allocation" is
97
a single allocation method to retain simplicity, we use the M1 allocation method to
calculate FCE.
Figure 4.4 depicts a forecast of the CO2 emissions from electricity generation
in Alberta under two scenarios and the corresponding emissions intensities. We
focus on the time period up to 2020, which coincides with the Provincial target of
50 MCO2 eq. of emissions reductions. This forecast considers the present genera
tion fleet, planned generation unit additions and retirements, and the new installed
capacity expected to meet the forecasted electricity demand to the year 20207. The
generation scenario GS1 assumes new additions that are yet unplanned will be coal
fired generators. Scenario GS2 considers an alternative case where these new ad
ditions will be cogeneration systems, employed in the oil sands sector. We assume
that carbon capture and storage will not be implemented within the time period
of this forecast. Both scenarios are plausible given the corporate announcements
made by utility companies to build new coal fired power plants and the forecasted
growth of oil sands sector combined with the potential to use cogeneration systems
to satisfy their energy demands (see Appendix E for details of the forecast). The
scenario GS1 is assumed as the business as usual (BAU) scenario due to the exist
ing large reserves of coal in Alberta, the potential to develop brownfield coal fired
generation to replace retiring units as well as the ability to expand the generation
capacity of existing coal fired generators. The transmission system expansions
announced by the AESO can facilitate either of these generation scenarios [46].
As shown in Figure 4.4, a 1117% reduction of Alberta electricity sector CO2
emissions below the BAU scenario could be achieved by integrating more cogen
eration. However, the use of GS1 as the BAU scenario is subject to challenge. A
suitable. However, in the case of cogeneration, this method is not suitable because the significantly
different qualities of the two energy products are not taken in to account7Planned additions are the units that are under active construction and the ones that have received
regulatory approval.
98
2001 2003 2005 2007 2009 2011 2013 2015 2017 201940
45
50
55
60
65
70
75
CO2 Emissions from Electricity Generation in Alberta
mill
ion
tCO
2
Actual Forecast
2001 2003 2005 2007 2009 2011 2013 2015 2017 20190.4
0.5
0.6
0.7
0.8
0.9
Average CO2 Emissions Intensity of the Alberta Electricty System
tCO
2/MW
h
year
Actual Forecast
GS2GS1GS2(M1)GS1(M1)
GS2GS1GS2(M1)GS1(M1)
Figure 4.4: Forecast of CO2 emissions from the Alberta electric system to 2020A forecast of CO2 emissions from the Alberta electric system to 2020 is presentedin this figure. The generation scenario GS1 is a high coal option and GS2 is ahigh cogeneration option (details of the two generation scenarios are summarized insection 4.4 and full details are presented in the Appendix E). The range of emissionsunder each scenario is due to the different allocation methods used to calculate theemissions intensity of cogenerated electricity. Therefore the range widens with theincreasing amount of cogenerated electricity in the mix. If the allocation methodM1 (incremental fuel based) is used to divide the fuel between steam and electricityproduced by a cogeneration system, the outlook of the total CO2 emissions (and theaverage CO2 intensity) attributable to the electricity generation in Alberta under thescenario GS1 and GS2 are depicted by the lines GS1(M1) and GS2(M1) respectively.Depending on the allocation method, the electricity sector emissions outlook underthe scenario GS2 (high cogeneration) is 1117% lower than that of GS1 (high coalor BAU).
99
2001 2003 2005 2007 2009 2011 2013 2015 2017 201940
45
50
55
60
65
70
75
CO2 Emissions from Electricity Generation in Alberta
mill
ion
tCO
2
Actual Forecast
2001 2003 2005 2007 2009 2011 2013 2015 2017 20190.4
0.5
0.6
0.7
0.8
0.9
Average CO2 Emissions Intensity of the Alberta Electricty System
tCO
2/MW
h
year
Actual Forecast
GS2GS3GS2(M1)GS3(M1)
GS2GS3GS3(M1)GS2(M1)
Figure 4.5: Forecast #2 of CO2 emissions from the Alberta electric system to 2020Generation scenario GS3 is a high CCGT BAU scenario, assuming that retiring coalunits will be replaced by CCGT units.
100
strict carbon emissions control regulation enacted by the province or the Canadian
federal government could constrain the growth of both the oil sands sector and coal
fired electricity generation. However, there is significant uncertainty in the timing
and stringency of such regulation. We test a third scenario (GS3) by assuming
that the new generation additions to replace the retiring units and to serve the
forecasted demand growth will be natural gas fired CCGT units (see Appendix E
for details). The high cogeneration scenario, GS2, is only 25% lower than the high
CCGT scenario, GS3, demonstrating that the choice of BAU significantly impacts
the estimates of the emissions reduction potential of cogeneration. It also sug
gests that a similar level of emissions reductions are possible through increased
deployment of natural gas fired CCGT generators.
There is significant risk in picking a technology winner as opposed to setting a
target standard that can be met using a mix or blend of technologies, each keyed to
the subregion or resource base being accessed. Therefore, we estimate the cost of
mitigating CO2 in the Alberta electricity sector using alternative electricity gener
ation technologies compared to a supercritical pulverized coal (SCPC) power plant
as shown in Table 4. SCPC was used as the new coal fired electricity generation
technology, as it is assumed to be the dominant technology of new coal fired units
that will be built before 2020. This is consistent with the new SCPC units that
are being built and are planned in Alberta [136]. However, the baseline chosen
for comparison will greatly affect these results and therefore, care should be taken
in selecting and interpreting the baseline for this type of analysis. The estimated
carbon mitigation cost of cogeneration compared to SCPC is 14 $/tCO2 (a negative
abatement cost means that under the assumed conditions, both the average cost
and the carbon intensity of cogenerated electricity are lower than SCPC), the low
est among the generation technologies considered. This carbon abatement cost is
101
Table 4.4: Cost of abated carbon emissionsEstimates of carbon mitigation costs of alternative electricity generation technologies compared to a supercritical pulverized coal power plant (‘‘Baseline" unit). In each case thetransmission costs are equally distributed across the grid and assumed to be built in proportionately to the power supplied. The Province has undertaken a series of transmissionupgrade projects sufficient to provide adequate future capacity to meet projected loads including oil sands expansion. Funding for right of way and capital costs will be apportionedinitially outside the rate base and charged back to reflect load served in operations.
SCPC CCGT CogenWindpower
Fuel Coal NG NG WindCapital cost ($/kW)a 3000 1365 1000 2200Fixed O&M cost ($/kWyear) 31 13 13 56Variable O&M cost ($/MWh) 6 4 4 0Fuel price ($/GJ) 1.5 6 6 0Fuel carbon intensity (tCO2/GJ) 0.1 0.05 0.05 0Heat rate (GJ/MWh)b 9.4 7.7 6.7 0Cost of electricity ($/MWh) 71 87 63 114c
Carbon intensity (tCO2/MWh) 0.94 0.39 0.34 0Cost of CO2 reduction ($/tCO2) Baseline 29 14 46
SCPC supercritical pulverized coal;
All costs are in 2008 Canadian dollars (average conversion rate in 2008 CAD 1=USD 0.94).a The source of capital costs of all generation technologies except cogeneration is [46]. Capital cost
of SCPC is based on a unit size of 450MW and that of CCGT is based on a unit size of 300MW. Co
generation capital cost attributable to electricity generation is assumed to be the difference between
the capital cost of a cogeneration system ( gas turbine + HRSG) and that of an industrial boiler with
identical steam generation capacity.b All heating values are based on higher heating values. Heat rate of the cogeneration unit is based
on the allocation method M1.c Cost of wind energy does not includes the cost of new transmission developments required to
integrate wind and the cost associated with mitigating the intermittency of wind.
102
lower than estimates for carbon capture and storage from new coal power plants,
which are in the range of $70100/tCO2 [58]. Given these results, cogeneration
presents an effective option to reduce the CO2 emissions of the Alberta electricity
sector.
Our analysis shows that, in general, the cogeneration option is economically
favourable compared to the baseline option. However, the economics of cogener
ation are tightly correlated with natural gas and electricity prices. With a natural
gas price of $5/GJ and an electricity price of $60/MWh, the total cost of energy
input per barrel of bitumen produced under the baseline option is $6.6 and that of
the cogeneration option is $5.5. The market price of electricity varies hour to hour
throughout the day because different generation units are dispatched to meet the
time varying electricity demand at the minimum cost. On average we expect the
hourly electricity price to be equal to the marginal cost of generation, which in turn
depends primarily on the fuel cost for thermal electricity generation.
We examine the competitiveness of cogenerated electricity under historic elec
tricity and natural gas prices in Alberta in order to determine the potential value
and role of cogeneration in the future. As discussed above we use the M1 alloca
tion method to calculate the marginal fuel consumption for cogenerated electricity.
Under the M1 method, the implied heat rate of the cogeneration system in our il
lustrative example system is 6.7GJ/MWh. The average annual natural gas price in
the years 2007 through 2009 in Alberta was $6.24/GJ, $7.81/GJ, and $3.93/GJ
respectively. The hourly electricity prices of the Alberta power market in those
years were less than the average fuel cost of cogenerated electricity 44%, 40% and
32% of the time respectively. We can also use the market heat rate8 to examine
the competitiveness of a generation technology under both electricity and natural
gas price fluctuations. In general, a generator with a heat rate above the prevailing
8Market heat rate=market price of electricity / natural gas price; expressed in GJ/MWh
103
market heat rate is operating at a loss. The heat rate of the cogeneration system
we model (6.7GJ/MWh; M1 allocation method) is higher than the hourly market
heat rate in Alberta in the years 2007 through 2009 47%, 46% and 28% of the time
respectively. Conventional thermal generating units such as CCGT can adjust their
output in response to these market fluctuations (e.g., reduce output when market
price is low and vice versa). However, cogeneration units typically follow the host
facility’s thermal load and cannot reduce or shut down electricity production fol
lowing the electricity price. Under these conditions, the economics from the power
sold by insitu extraction projects is not always favourable so they may choose to
size power generation capacity to meet their own needs rather than sell to the grid.
4.4.2 Policy Implications
In order to determine whether the current Alberta policy is sufficient to incent
investments in cogeneration, we calculate the emissions reduction obligations of
the two options under SGER according to the guidelines set by Alberta Environ
ment [138]. Results of SGER obligations calculations are shown in Figure 4.6 (see
Appendix C for SGER obligations calculations details). The baseline option has an
annual emissions reduction obligation of 63,000 tCO2 and the cogeneration option
earns 15,000 tCO2 of EPCs. As discussed in section 4.2.3, the present Alberta
GHG emissions reduction policy implicitly caps the price of carbon at $15/tCO2.
Therefore, the SGER compliance cost of the baseline option is $0.1/bbl of bitu
men. For perspective, if this was factored into energy of this option, the energy cost
would increase by 1.5%. In case of the cogeneration option the EPCs earned under
SGER translates to a savings of $0.02/bbl of bitumen, reducing the energy cost
only by 0.4%. If the value of EPCs earned under SGER is attributed to electricity,
the marginal cost of cogenerated electricity will reduce by $0.34/MWh. As men
104
M1 M2 M3 M4 ABav ABom−300
−250
−200
−150
−100
−50
0
50
100
150E
mis
sion
s re
duct
ion
oblig
atio
ns, (
1000
tCO
2/ye
ar)
SGERobligations(baselineoption)
SGERobligations
(cognerationoption)
Modified SGER obligations calculations for thecogeneration option
Deemed emissions attributed toelectricty is calculated using theemissions intensity of cogeneratedelectricity
Deemed emissions attributedto electricty is calculatedusing grid emissionsintensities
Figure 4.6: Emissions reductions obligationsThe first two columns of this chart depict the emissions reduction obligations of thebaseline option and the cogeneration option under the current SGER rules. Nextfour columns depict the emissions reduction obligations calculated with modifiedSGER guidelines where the emissions intensity of cogenerated electricity underdifferent allocation methods (M1M4) is used to calculate the deemed emissionsfrom electricity instead of the CCGT emissions intensity. This modification to thepresent SGER rules creates an unfavourable situation for the cogeneration optioneither by obligating to reduce emissions or by reducing EPCs. However, underall allocation methods, except M2 method (energy based), the cogeneration optionis still the preferred option in terms of emissions reduction obligations. The lasttwo columns depict the amount of EPCs the cogeneration option under SGER ifAlberta grid intensities (average and marginal intensities) are used to calculatethe deemed emissions attributed to electricity. As can be seen from the figure,such modifications to SGER rules create a favourable environment for cogenerationoption.
105
tioned above, without SGER benefits, the marginal cost9 of cogenerated electricity
was higher than the electricity prices in Alberta in 2008 and 2009 40% and 32% of
the time respectively. Lowered marginal cost due to the SGER performance credits
of $0.34/MWh reduces the fraction of time where the marginal cost is higher than
the electricity price less than one percentage point in both years (we consider only
2008 and 2009 because the SGER compliance period started in 2008). Hence, the
current Alberta GHG emissions reduction regulation in its present form is not suf
ficient to considerably increase the competitiveness of cogeneration and influence
cogeneration investment decision making.
Another limitation of SGER is the use of CO2 emissions intensity of a CCGT
unit to calculate the ‘‘deemed emissions attributed to electricity" as described in
section 4.2.3. In this case the SGER guidelines assume that in the absence of co
generation systems, the electricity demand of the host facility will be met by CCGT
units. Given the present generation mix in Alberta and new generation additions
that either have regulatory approval or are under active construction, this is not
a realistic assumption [136]. Under the present regulatory environment, coal is
still likely to be the dominant generation technology, which will result in a high
average electricity emissions intensity. Instead of using the CO2 intensity of CCGT
(to calculate the ‘‘deemed emissions attributed to electricity"), one of the alloca
tion methods could be used. However, this would create a worse off situation for
cogeneration, either by increasing the emissions reduction obligations (allocation
methods M1M3) or by reducing the amount of EPCs that may be earned compared
to EPCs earned under current SGER rules (see Figure 4.6).
When there is a significant amount of cogeneration in the electricity generation
mix, the emissions intensity of cogenerated electricity, Icogen is required to calcu
9Marginal cost is assumed to be equal to the sum of fuel cost and variable O&M costs
106
late both the average and the marginal CO2 emissions intensity10. However, as
described in section 4.3, Icogen depends on the allocation method (i.e., how the
emissions are divided between electricity and heat/steam; see Figure 4.3) and
therefore, the method employed affects the average and marginal CO2 emissions
intensity. For example, as shown in Figure 4.4 the exact value of the total CO2
emissions and the average emissions intensity of the Alberta electric sector de
pend on the allocation method used to calculate Icogen. It can also be seen that
the range widens with the increasing share of cogenerated electricity (in 2009 the
variability in total CO2 emissions depending on the allocation method employed
was 5.6 MtCO2). Therefore, a carbon management policy that uses the average or
marginal emissions intensities of the electric system must also set the allocation
method that should be used to calculate the emissions intensity of cogeneration
units. Furthermore, different cogeneration system configurations (steam turbine
based, gas turbine based etc.) that are/could be employed complicate the esti
mation of emissions intensities by using aggregated data. For simplicity, when
preparing the emissions forecast depicted in Figure 4.4, we apply the Icogen values
(see Figure 4.3) from our model to all the cogeneration units in the Alberta genera
tion mix. Through sensitivity analysis we are confident that the values we use are
of the same order of the magnitude of the emissions intensities of the respective
cogeneration units under the allocation methods M1M4. A comprehensive survey
of cogeneration units employed in the generation mix is required to make a more
accurate estimate of associated emissions intensities.
10In Alberta currently about 30% of the electricity is generated by cogeneration units while they
operate in the margin (ie. set the price) 25% of the time on average [153,154]
107
Carbon free electricity M1 M2 M3 M4
0
100
200
300
400
500
600
700
Em
issi
ons
offs
et c
redi
ts, (
ktC
O2/y
ear)
Allocation method
Displace grid average (S1)Displace grid OM (S2)Displace coal (S3)Displace CCGT (S4)
Figure 4.7: Emissions offset creditsCO2 emissions offset credits that may be earned by the cogeneration option aredepicted in this figure. The amount of credits depends on two factors: the allocation method used to calculate the emissions intensity of cogenerated electricityand the emissions intensity of displaced electricity. This figure shows the offsetcredits under the four allocation methods we considered (M1M4) and four displacement scenarios (S1S4). The group ‘‘carbon free electricity" shows the offsetcredits earned by a carbon free electricity generation unit (such as wind power,photovoltaics, biomass etc.) under the four displacement scenarios and is shownfor comparison. This may also viewed as the offset credits earned by the cogeneration system if all the emissions are allocated to steam and electricity is consideredto be emissions free.
4.4.3 Policy Options
We explore alternate policy options and their ability to increase the competitiveness
of cogeneration. First, we consider a case where the carbon management policy
allows the cogeneration systems to earn carbon emissions offset credits for grid
electricity displacements. Annual offset credits that our modeled system may earn
under different allocation methods (M1M4) and different electricity offset scenarios
(S1S4) are shown in Figure 4.7. These credits are calculated using equation (4.11)
108
as described in section 4.3. A comparison of Figures 4.6 and 4.7 shows that
all the offset scenarios except S4 with the allocation method M4 provides higher
credits for the cogeneration system than SGER EPCs. These offset credits may be
used to meet the facility’s own emissions reduction obligations or be sold to other
parties who have emissions reduction obligations. An Alberta based offset credits
market already exists to sell credits for parties who have SGER emissions reduction
obligations.
It is also possible to provide more credits to the facilities with cogeneration
within the SGER framework by changing the method used to calculate the deemed
emissions attributed to electricity. Instead of using the emissions intensity of
CCGT, as is the case of the current procedure, the average emissions intensity of
the Alberta electricity sector may be used. This would represent the case where
cogenerated electricity displaces the average generation mix, which is dominated
by coal fired generation. Use of the current average emissions intensity of 0.84
tCO2/MWh as the basis of calculating the deemed emissions attributed to electricity
would increase the EPCs earned by the cogeneration option to 292,000 tCO2 from
15,000 tCO2 under the current guidelines (see Figure 4.6). Attributing all the EPCs
earned under this modified SGER obligation calculation to electricity at $15/tCO2
reduces the marginal cost of cogenerated electricity by $6.7/MWh. Similarly, if a
cogeneration system operator participates in the electricity market by following load
(instead of following their own thermal demands) the marginal emissions intensity
of the Alberta electricity sector could be used to calculate the deemed emissions
attributed to electricity. These conditions result in a significant benefit for facilities
with cogeneration.
As discussed above when controlling carbon emissions from multiproduct fa
cilities such as cogeneration through regulations based on offset credits for lower
109
carbon intensive technologies, or facility based intensity reduction targets such as
the SGER, the regulator is faced with the challenge of selecting the appropriate
method to allocate a facility’s emissions among multiple outputs. Furthermore,
in the case of offset credits based systems, particularly electricity offsets, there is
significant uncertainty in determining what is being displaced by the low carbon
alternative. This fact merit further analysis. For example, if the assumption is that
cogenerated electricity displaces a single type of generation technology such as coal
or CCGT (Figure 4.3 & 4.7; scenarios S3S4), the CO2 intensity of a representative
unit of that technology should be determined at the time of policy adoption. That
decision should be made considering the existing generating units as well as future
generation unit additions. Of the four offset scenarios considered in this analysis,
the required information to calculate the grid average emissions (scenario S1) in
tensity may be already available from various emissions reporting sources. For
example, Alberta’s ‘‘Specified Gas Reporting Regulation’’ requires the major CO2
emitters such as electric power producers to report their emissions annually [155].
Nevertheless, considerable uncertainty remains as to the accuracy of the assump
tion the displaced electricity emissions intensity is equal to the average grid inten
sity. Compared to other emissions intensities, the marginal emissions intensity,
which is required to calculate offset credits under scenario S2, is the most difficult
to calculate with reasonable certainty. In order to calculate the marginal intensity
the regulator must know which generating unit was operating at the margin over
a given time frame as well as its emissions intensity. In a deregulated market en
vironment such information is privileged and only the independent electric system
operator (in Alberta the AESO) has the full knowledge of the marginal unit. Various
aggregated data sources are available (for example, [153] and [154]), although the
accuracy of the marginal emissions intensity derived from them is debatable.
110
Figures 4.6 and 4.7 depict the uncertainties in the incentives or obligations
for the cogeneration system in our model due to different allocation methods and
electricity displacement scenarios. If the regulator chooses to implement carbon
pricing by using facility or product based regulations, the emissions accounting
methods must be chosen in such a way that they match the intended policy objec
tives. For example, consider the results presented in Figure 4.6. If the objective of
the policy is to provide a significant amount of credits for cogeneration to promote
investment, the SGER rules may be modified, such that the deemed emissions at
tributed to electricity is calculated using grid average intensity. Conversely, if the
policy maker wishes to promote low carbon emissions intensive operations without
giving as many credits as the current SGER rules, the deemed emissions attributed
to electricity may be calculated using the emissions intensity of cogenerated elec
tricity under M1 allocation method. In this case no net credits are granted to a
bitumen extraction project with cogeneration, yet its emissions reduction obliga
tions are lower than that of a project without cogeneration.
4.5 Conclusions
Oil sands operations will likely provide a significant share of crude oil deliveries
within North America for the next few decades, with corresponding demand for
natural gas and delivered electricity to support their operations. Use of cogener
ation to satisfy the energy demands of oil sands operations may be an effective
strategy for reducing CO2 emissions of the electricity sector of Alberta. However,
this conclusion is likely to be true and most effective in the short run (before 2020)
when installed coal generation with limited emissions controls continues to supply
a significant fraction of electricity in the province. Beyond this point, it is likely that
displacement of electricity generated from natural gas (and other lower emissions
111
intensity sources) may offset or diminish the value of cogeneration for carbon man
agement in Alberta. In the face of this trend, with falling electric sector emissions,
long term oilsand cogeneration benefits may be most effective and sustaining if
installed immediately.
Cogeneration can offset a significant and locationally important segment of Al
berta’s base load electricity demand currently satisfied by coal fired generators. The
regulatory system can facilitate the integration of cogeneration systems within oil
sands operations through a combination of permits, tax incentives and regulatory
credits. The result in the short term will be measurable benefits from emissions re
ductions associated with the electricity sector. However, since the present carbon
management policy of Alberta does not impose a significant marginal carbon price
signal there is limited influence on oil sands project operator’s decisions to invest
in cogeneration. With a strong carbon price signal, cogenerated electricity will be
a more competitive base load generation option.
A more efficient solution is available, simply by focusing on a carbon tax. Here,
the fuel used can be taxed based on its carbon intensity, resulting in an economy
wide, consistent carbon price. Use of a lower carbon intensive fuel such as natural
gas combined with the inherently high efficiency will make cogeneration competitive
compared to other electricity generation technologies (see Table 4). Furthermore,
enforcing a price on carbon at the source eliminates the need for down stream car
bon accounting that demands significant data collection and complex accounting
methods.
When facing a lack of political will for a carbon tax, alternative methods should
be chosen to mimic the effect of such a tax. This merits further research. For
example with respect to cogeneration, future work could provide guidance on the
accounting methods such as coproduct allocation that provide the same level of
112
incentives as a carbon tax.
We may draw more general lessons from this analysis. Regulations that at
tempt to manage emissions on a product and facility basis may become arbitrary
and complex as regulators attempt to approximate the effect of an economywide
carbon price. If one counts only the direct emissions from facilities, then the sys
tem is simple, but encourages counterproductive activity as industry might try to
move emissions outside their ‘‘fence". Though less supported in the current po
litical climate, economywide policies would address offsite emissions in a more
direct manner. Regulators can attempt to improve the regulations by accounting
for indirect emissions on a product basis, in this case emissions from purchased
electricity, to avoid such perverse outcomes. But as one adds more complexity the
system becomes more arbitrary, and more subject to gaming by industry.
Improvements to the transparency of carbon management policies include clearly
stating the methods for accounting procedures and assumptions made. In addi
tion, all the data associated with calculating emissions of a product or a facility
should be made easily accessible in the public domain. As demonstrated in this
analysis, a number of rational emissions accounting methods are available and
they provide different levels of incentives for cogeneration. Therefore, policy mak
ers should select the appropriate accounting methods that reflect the intended
policy goals.
113
Chapter 5
Conclusions
This thesis explored the effectiveness of wind power and natural gas fired cogener
ation for carbon management of electric power systems using the Alberta electric
power system as a case study. The results of this work are intended to support
efficient climate change mitigation policy making. While the principal focus is
carbon management of electric power systems, this thesis also examined strate
gic solutions for the challenges faced by a wind power producer participating in
conventional electricity markets.
Chapter 2 assessed the effectiveness of wind power for carbon emissions man
agement of electric power systems. Operations of Alberta electric power system
with wind penetration levels of 060% were simulated in order to assess the carbon
emissions abatement potential of wind power. The main contribution of this work
is a set of carbon abatement supply curves. Policy makers can use these supply
curves to weigh the competitiveness of wind power for carbon management against
other available options. In this chapter, it was shown that after accounting for the
costs and CO2 emissions incurred in mitigating variability, wind power has the po
tential to abate 216 million tCO2/year in Alberta at a marginal abatement cost of
110120 $/tCO2 (in 2010 Canadian dollars). In the Alberta electric power system,
under the assumed conditions, the cost of wind power variability and uncertainty
was a modest 14 $/MWh of wind power at wind power penetration levels of 560%.
In many jurisdictions electric power generation is a competitive business. There
fore, the market competitiveness of wind power is important to attract investments.
Due to the variability and uncertainty of wind power, two factors—potential low
114
correlation of wind power production with demand and having to settle energy
imbalances—may lead to lower profits for a WPP participating in conventional elec
tricity markets. Chapter 3 explored the use of CAES systems to mitigate those
challenges. The main contribution of this chapter is the development of a model
that can be used to assist optimal operations decisions of a WPP and a CAES sys
tem that jointly participate in a dayahead electricity market. The model inherently
takes the uncertainty in future wind power availability and price of electricity by
utilizing a twostage stochastic programming approach. The main results of the
model are a set of robust bids for a dayahead electricity market and a set of CAES
system operating rules. Results of a case study conducted using the model showed
that integrating a CAES system with a WPP is an effective strategy for increasing the
economic value of wind power, managing energy imbalances, and managing finan
cial risks. The advantage of taking a stochastic programming approach for decision
modeling under uncertainty was demonstrated by quantifying the expected value
of perfect information.
Chapter 4 assessed the role of cogeneration for managing carbon emissions in
Alberta. The analysis is extended to evaluate the effectiveness of Alberta’s cur
rent and alternative carbon emissions control regulations. The results suggest
that the use of cogeneration for satisfying the energy demands of the oil sands
operations in Alberta is an effective option for reducing carbon emissions from Al
berta’s electricity sector. However, the long term emissions reduction benefits of
oilsand cogeneration may be most effective and sustained if installed immediately.
This analysis showed that the current emissions control regulation of Alberta does
not create a strong marginal carbon price that can influence the investments on
low carbon intensive electricity generation technologies such as cogeneration. By
taking oilsands cogeneration as an example case, the analysis provided policy in
115
sights that illustrate how the choice of accounting methods may complicate the
implementation of facility or productbased carbon emissions control regulations
such as Alberta’s current regulation. The analysis results showed that the different
accounting methods and calculations of electricity offsets could lead to very differ
ent levels of incentives for cogeneration. We conclude that the choice of emissions
accounting methods for facility or productbased regulations should involve the
policy maker and be driven by policy goals.
Bibliography
[1] IPCC, Climate Change 2007: The Physical Science Basis. Contribution of Work
ing Group I to the Fourth Assessment Report of the Intergovernmental Panel
on Climate Change, S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis,K. B. Averyt, M. Tignor, and H. L. Miller, Eds. Cambridge, UK and New York,USA: Cambridge University Press, 2007.
[2] J. Houghton, Global Warming: The Complete Briefing, 4th ed. Cambridge,UK: Cambridge University Press, 2009.
[3] J. G. Canadell, C. Le Quere, M. R. Raupach, C. B. Field, E. T. Buitenhuis,P. Ciais, T. J. Conway, N. P. Gillett, R. A. Houghton, and G. Marland, ‘‘Contributions to accelerating atmospheric CO2 growth from economic activity,carbon intensity, and efficiency of natural sinks,’’ Proceedings of the National
Academy of Sciences, vol. 104, no. 47, pp. 18 866–18870, 2007.
[4] G. Heal and B. Kristrom, ‘‘Uncertainty and climate change,’’ Environmental
and Resource Economics, vol. 22, no. 1, pp. 3–39, 2002.
[5] J. Rockstrom, W. Steffen, K. Noone, Å. Persson, F. S. Chapin, E. F. Lambin,T. M. Lenton, M. Scheffer, C. Folke, H. J. Schellnhuber, B. Nykvist, C. A.de Wit, T. Hughes, S. van der Leeuw, H. Rodhe, S. Sorlin, P. K. Snyder,R. Costanza, U. Svedin, M. Falkenmark, L. Karlberg, R. W. Corell, V. J.Fabry, J. Hansen, B. Walker, D. Liverman, K. Richardson, P. Crutzen, andJ. A. Foley, ‘‘A safe operating space for humanity,’’ Nature, vol. 461, no. 7263,pp. 472–475, Sep. 2009.
[6] M. Meinshausen, N. Meinshausen, W. Hare, S. C. B. Raper, K. Frieler,R. Knutti, D. J. Frame, and M. R. Allen, ‘‘Greenhousegas emission targetsfor limiting global warming to 2◦C,’’ Nature, vol. 458, no. 7242, 2009.
[7] K. Caldeira, A. K. Jain, and M. I. Hoffert, ‘‘Climate Sensitivity Uncertaintyand the Need for Energy Without CO2 Emission,’’ Science, vol. 299, no. 5615,pp. 2052–2054, 2003.
[8] IEA, Key World Energy Statistics 2010. Paris: International Energy Agency,2010.
[9] J. Apt, D. W. Keith, and M. G. Morgan, ‘‘Promoting LowCarbon ElectricityProduction,’’ Issues in Science and Technology, vol. 23, no. 3, pp. 37–42,2007.
[10] IEA, Energy Technology Perspectives: Scenarios & Strategies to 2050. Paris:International Energy Agency (IEA), 2008.
116
117
[11] S. Pacala and R. Socolow, ‘‘Stabilization Wedges: Solving the Climate Problemfor the Next 50 Years with Current Technologies,’’ Science, vol. 305, no. 5686,pp. 968–972, 2004.
[12] M. I. Hoffert, ‘‘Advanced Technology Paths to Global Climate Stability: Energyfor a Greenhouse Planet,’’ Science, vol. 298, no. 5595, pp. 981–987, 2002.
[13] GWEC, ‘‘Global wind report: Annual market update 2010,’’Global Wind Energy Council, Tech. Rep., 2011. [Online]. Available:http://www.gwec.net/index.php?id=180
[14] CanWEA. (2011) Canadian wind farms. [Online]. Available:http://www.canwea.ca/farms/index_e.php
[15] J. G. McGowan and S. R. Connors, ‘‘Windpower: A turn of the century review,’’ Annual Review of Energy and the Environment, vol. 25, no. 1, pp.147–197, 2000.
[16] R. Wiser and M. Bolinger, ‘‘2008 wind technologies market report,’’U.S. Department of Energy, Tech. Rep., 2009. [Online]. Available:http://www.windpoweringamerica.gov/filter_detail.asp?itemid=2788
[17] L. Bird, M. Bolinger, T. Gagliano, R. Wiser, M. Brown, and B. Parsons, ‘‘Policies and market factors driving wind power development in the united states,’’Energy Policy, vol. 33, no. 11, pp. 1397–1407, 2005.
[18] I. H. Rowlands, ‘‘The European directive on renewable electricity: conflictsand compromises,’’ Energy Policy, vol. 33, no. 8, pp. 965–974, 2005.
[19] P. Siemes, H.J. Haubrich, H. Vennegeerts, and S. Ohrem, ‘‘Concepts for theimproved integration of wind power into the german interconnected system,’’IET Renewable Power Generation, vol. 2, no. 1, pp. 26 –33, 2008.
[20] E. Martinot, C. Dienst, L. Weiliang, and C. Qimin, ‘‘Renewable energy futures: Targets, scenarios, and pathways,’’ Annual Review of Environment
and Resources, vol. 32, no. 1, pp. 205–239, 2007.
[21] DOE, ‘‘20% Wind Energy by 2030: Increasing Wind Energy’s Contribution to U.S. Electricity Supply,’’ U.S. Department of Energy, DOE/GO1020082567, 2008. [Online]. Available:http://www1.eere.energy.gov/windandhydro/pdfs/41869.pdf
[22] EREC, ‘‘Renewable Energy Technology Roadmap: 20% by2020,’’ European Renewable Energy Council, Tech. Rep.,2008. [Online]. Available: http://www.erec.org/fileadmin/erec_docs/Documents/Publications/Renewable_Energy_Technology_Roadmap.pdf
118
[23] J. F. DeCarolis and D. W. Keith, ‘‘The Economics of Largescale Wind Powerin a Carbon Constrained World,’’ Energy Policy, vol. 34, no. 4, pp. 395 – 410,2006.
[24] J. Smith, M. Milligan, E. DeMeo, and B. Parsons, ‘‘Utility wind integrationand operating impact state of the art,’’ IEEE Transactions on Power Systems,vol. 22, no. 3, pp. 900 –908, 2007.
[25] W. Grant, D. Edelson, J. Dumas, J. Zack, M. Ahlstrom, J. Kehler, P. Storck,J. Lerner, K. Parks, and C. Finley, ‘‘Change in the air,’’ IEEE Power and
Energy Magazine, vol. 7, no. 6, pp. 47–58, 2009.
[26] J. Apt, ‘‘The spectrum of power from wind turbines,’’ Journal of Power
Sources, vol. 169, no. 2, pp. 369–374, 2007.
[27] E. Denny and M. O’Malley, ‘‘Quantifying the total net benefits of grid integrated wind,’’ IEEE Transactions on Power Systems, vol. 22, no. 2, pp. 605–615, 2007.
[28] H. Holttinen, ‘‘Estimating the impacts of wind power on power systems—summary of IEA Wind collaboration,’’ Environmental Research Letters, vol. 3,no. 2, p. 025001, 2008.
[29] T. Ackermann, J. Abbad, I. Dudurych, I. Erlich, H. Holttinen, J. Kristoffersen, and P. Sorensen, ‘‘European Balancing Act,’’ IEEE Power and Energy
Magazine, vol. 5, no. 6, pp. 90–103, 2007.
[30] GE Energy, ‘‘Ontario wind integration study,’’ GE Energy forOntario Power Authority, Tech. Rep., 2006. [Online]. Available:http://www.ieso.ca/imoweb/pubs/marketreports/OPAReport2006101.pdf
[31] AESO, ‘‘Incremental Impact on System Operations with Increased Wind Power Penetration Phase 1,’’ Alberta Electric System Operator, Tech. Rep., 2005. [Online]. Available: http://www.aeso.ca/downloads/Incremental_Effects_on_System_Operations_with_Increased_Wind_Power_Penetration_rev_2_3.pdf
[32] ——, ‘‘Assessing the impacts of increased wind power on AIES operationsand mitigating measuresPhase 2,’’ Alberta Electric System Operator, Tech.Rep., 2006. [Online]. Available: http://www.aeso.ca/downloads/AESO_Phase_II___Wind_Integration_Impact_Studies_final_20060718.pdf
[33] AESO, ‘‘Phase Two Wind Integration,’’ Alberta Electric System Operator, Calgary, Tech. Rep., 2010. [Online]. Available:http://www.aeso.ca/gridoperations/22412.html
[34] H. Holttinen, M. Milligan, B. Kirby, T. Acker, V. Neimane, and T. Molinski,‘‘Using Standard Deviation as a Measure of Increased Operational Reserve
119
Requirement for Wind Power,’’ Wind Engineering, vol. 32, no. 4, pp. 355–377,2009.
[35] J. Greenblatt, S. Succar, D. Denkenberger, R. Williams, and R. Socolow,‘‘Baseload wind energy: modeling the competition between gas turbines andcompressed air energy storage for supplemental generation,’’ Energy Policy,vol. 35, no. 3, pp. 1474–1492, 2007.
[36] L. Dale, D. Milborrow, R. Slark, and G. Strbac, ‘‘Total cost estimates for largescale wind scenarios in UK,’’ Energy Policy, vol. 32, no. 17, pp. 1949–1956,2004.
[37] GWEC, ‘‘Global wind energy outlook 2008,’’ Global Wind Energy Council,Brussels and Greenpeace, Amsterdam,, Tech. Rep., 2008.
[38] W. Katzenstein and J. Apt, ‘‘Air emissions due to wind and solar power,’’Environmental Science & Technology, vol. 43, no. 2, pp. 253–258, 2009, pMID:19238948.
[39] H. Holttinen and S. Tuhkanen, ‘‘The effect of wind power on CO2 abatementin the Nordic Countries,’’ Energy Policy, vol. 32, no. 14, pp. 1639–1652, 2004.
[40] M. Hoogwijk, D. van Vuuren, B. de Vries, and W. Turkenburg, ‘‘Exploring theimpact on cost and electricity production of high penetration levels of intermittent electricity in OECD Europe and the USA, results for wind energy,’’Energy, vol. 32, no. 8, pp. 1381–1402, 2007.
[41] A. J. Wood and B. F. Wollenberg, Power Generation Operation and Control,2nd ed. New York: John Wiley & Sons Inc., 1996.
[42] R. Sioshansi and W. Short, ‘‘Evaluating the Impacts of RealTime Pricing onthe Usage of Wind Generation,’’ IEEE Transactions on Power Systems, vol. 24,no. 2, pp. 516–524, 2009.
[43] J. Wang, A. Botterud, and G. Conzelmann, ‘‘Impact of Wind Power Forecasting on Unit Commitment and Dispatch,’’ in 8th International Workshop on
Largescale Integration of Wind Power into Power Systems, Bremen, 2009.
[44] AESO, ‘‘Longterm transmission system plan 2011,’’ AlbertaElectric System Operator, Tech. Rep., 2011. [Online]. Available:http://www.aeso.ca/transmission/22021.html
[45] Environment Canada, ‘‘National inventory report: Greenhouse gas sourcesand sinks in Canada 19902008,’’ Environment Canada, Tech. Rep., 2010.[Online]. Available: http://www.ec.gc.ca/Publications/default.asp?lang=En
[46] AESO, ‘‘Longterm transmission system plan 2009,’’ AlbertaElectric System Operator, Tech. Rep., 2009. [Online]. Available:http://www.aeso.ca/downloads/AESO_LTTSP_Final_July_2009.pdf
120
[47] AESO. (2011) Wind Power / AIL Data. [Online]. Available:http://www.aeso.ca/gridoperations/20544.html
[48] D. Miller and M. Chorel, ‘‘A Study on the Efficiency of Alberta’s ElectricalSupply System,’’ JEM Energy & Associates, Tech. Rep., 2004. [Online].Available: http://www.hme.ca/reports/
[49] C. Grigg, P. Wong, P. Albrecht, R. Allan, M. Bhavaraju, R. Billinton, Q. Chen,C. Fong, S. Haddad, S. Kuruganty, W. Li, R. Mukerji, D. Patton, N. Rau,D. Reppen, A. Schneider, M. Shahidehpour, and C. Singh, ‘‘The IEEE Reliability Test System1996. A report prepared by the Reliability Test SystemTask Force of the Application of Probability Methods Subcommittee,’’ IEEE
Transactions on Power Systems, vol. 14, no. 3, pp. 1010–1020, 1999.
[50] J. Ihle and B. Owens, ‘‘Integrating Coal and Wind PowerDevelopment in the U.S. Upper Great Plains,’’ Platts Research Consulting, Tech. Rep., 2004. [Online]. Available:http://www.dora.state.co.us/puc/projects/ReliableInfrastructure/TFResources/IntegratingCoalPlatts.pdf
[51] IEA, ‘‘Power Generation from Coal,’’ International Energy Agency, Paris, Tech.Rep., 2010. [Online]. Available: http://www.iea.org/ciab/papers/power_generation_from_coal.pdf
[52] J. B. Klein, ‘‘The Use of Heat rates in Production Cost modeling and MarketModeling,’’ California Energy Commission, Tech. Rep., 1998. [Online].Available: http://nodal.ercot.com/docs/tntarc/mo/CAHeatRates.pdf
[53] J. R. MacCormack, H. Zareipour, and W. D. Rosehart, ‘‘A Reduced Model ofthe Alberta Electric System for Policy, Regulatory, and Future DevelopmentStudies,’’ in IEEE Power and Energy Society General Meeting, Pittsburgh, PA,2008.
[54] M. Blanchard and G. Desrochers, ‘‘Generation of autocorrelated wind speedsfor wind energy conversion system studies,’’ Solar Energy, vol. 33, no. 6, pp.571–579, 1984.
[55] J. R. MacCormack, D. Westwick, H. Zareipour, and W. D. Rosehart, ‘‘Stochastic modelling of future wind generation scenarios,’’ in 40th IEEE North Amer
ican Power Symposium (NAPS), 2008.
[56] EIA, ‘‘Voluntary Reporting of Greenhouse Gases: Instructions for FormEIA1605,’’ Energy Information Administration, Tech. Rep., 2009. [Online].Available: http://www.eia.doe.gov/oiaf/1605/pdf/EIA1605_Instructions_102307.pdf
[57] Alberta Energy. (2011) Carbon capture and storage. [Online]. Available:http://www.energy.alberta.ca/Initiatives/1438.asp
121
[58] ICON, ‘‘Carbon dioxide capture and storage: A canadian clean energyopportunity,’’ Integrated CO2 Network, Tech. Rep., 2009. [Online]. Available:http://www.ico2n.com/wpcontent/uploads/2010/07/ICO2NReport_09_final2.pdf
[59] S. Lefton, ‘‘The Cost of Cycling Coal Fired Power Plants,’’ Coal Power Maga
zine, 2006.
[60] J. Gostling, ‘‘Two Shifting of Power Plant: Damage to Power Plant Due toCycling A brief overview,’’ OMMI, vol. 1, no. 1, 2002.
[61] IEEE PES Power Syst Eng Comm, ‘‘Operational aspects of generation cycling,’’ IEEE Transactions on Power Systems, vol. 5, no. 4, pp. 1194–1203,1990.
[62] ERCB, ‘‘Alberta’s energy reserves 2008 and supply/demand outlook 20092018,’’ Energy Resources and Conservation Board, Tech. Rep. ST98, 2009.[Online]. Available: http://www.ercb.ca/docs/products/STs/st982009.pdf
[63] J. Chang, K. Madjarov, R. Baldick, A. Alvarez, and P. Q. Hanser, ‘‘Renewableintegration model and analysis,’’ in 2010 IEEE PES Transmission and Dis
tribution Conference and Exposition. New Orleans: IEEE, 2010, statisticalmodelling of regulation and load following costs incurring from renewablepower integration.
[64] WWEA, ‘‘World Wind Energy Report 2010,’’ World Wind Energy Association, Bonn, Tech. Rep., 2011. [Online]. Available:http://www.wwindea.org/home/images/stories/pdfs/worldwindenergyreport2010_s.pdf
[65] (2011) Wind Powering America: U.S. Installed Wind Capacity and Wind Project Locations. [Online]. Available:http://www.windpoweringamerica.gov/wind_installed_capacity.asp
[66] EIA, ‘‘Electric Power Annual,’’ Energy Information Administration, Tech.Rep., 2010.
[67] REN, ‘‘Renewables 2010 Global Status Report,’’ REN21 Secretariat, Paris,Status report, 2010. [Online]. Available: http://www.ren21.net/
[68] R. Wiser, M. Bolinger, and G. Barbose, ‘‘Using the Federal Production TaxCredit to Build a Durable Market for Wind Power in the United States,’’ The
Electricity Journal, vol. 20, no. 9, pp. 77–88, 2007.
[69] J. I. Lewis and R. H. Wiser, ‘‘Fostering a renewable energy technology industry: An international comparison of wind industry policy support mechanisms,’’ Energy Policy, vol. 35, no. 3, pp. 1844–1857, 2007.
122
[70] T. Ackermann, G. Ancell, L. Borup, P. Eriksen, B. Ernst, F. Groome,M. Lange, C. Mohrlen, A. Orths, J. O’Sullivan, and M. de la Torre, ‘‘Wherethe wind blows,’’ IEEE Power and Energy Magazine, vol. 7, no. 6, pp. 65–75,2009.
[71] S. Fink, J. Rogers, and K. Porter. (2009) Wind Power and Electricity Markets.[Online]. Available: http://www.uwig.org/windinmarketstableSep07.pdf
[72] S. Stoft, Power System Economics: Designing Markets for Electricity. IEEEPress, 2002.
[73] G. Rothwell and T. Gomez, Electricity Economics: Regulation and Deregula
tion. Piscataway,NJ: IEEE Press, 2003.
[74] C. Monteiro, R. Bessa, V. Miranda, A. Botterud, J. Wang, andG. Conzelmann, ‘‘Wind Power Forecasting: StateoftheArt 2009,’’ ArgonneNational Laboratory, Argonne, IL, Tech. Rep., 2009. [Online]. Available:http://www.dis.anl.gov/pubs/65613.pdf
[75] B. Ernst, B. Oakleaf, M. Ahlstrom, M. Lange, C. Moehrlen, B. Lange,U. Focken, and K. Rohrig, ‘‘Predicting the Wind,’’ IEEE Power and Energy
Magazine, vol. 5, no. 6, pp. 78–89, 2007.
[76] J. GarcıaGonzalez, R. de La Muela, L. Santos, and A. Gonzalez, ‘‘Stochastic Joint Optimization of Wind Generation and PumpedStorage Units in anElectricity Market,’’ IEEE Transactions on Power Systems, vol. 23, no. 2, pp.460–468, 2008.
[77] H. Holttinen and G. Koreneff, ‘‘Imbalance costs of wind power for a hydropower producer in Finland,’’ in Proc. European Wind Energy Conference 2007,2007, pp. 7–10.
[78] M. Robinson, ‘‘Role of balancing markets in wind integration,’’ in 2006 IEEE
PES Power Systems Conference and Exposition, 2006, pp. 232 –233.
[79] R. Sioshansi, ‘‘Increasing the Value of Wind with Energy Storage,’’ The Energy
Journal, vol. 32, no. 2, 2011.
[80] P. Twomey and K. Neuhoff, ‘‘Wind power and market power in competitivemarkets,’’ Energy Policy, vol. 38, no. 7, pp. 3198–3210, 2010.
[81] H. Holttinen, ‘‘Optimal electricity market for wind power,’’ Energy Policy,vol. 33, no. 16, pp. 2052–2063, 2005.
[82] L. Vandezande, L. Meeus, R. Belmans, M. Saguan, and J.M. Glachant, ‘‘Wellfunctioning balancing markets: A prerequisite for wind power integration,’’Energy Policy, vol. 38, no. 7, pp. 3146 – 3154, 2010.
[83] B. Ummels, M. Gibescu, and W. Kling, ‘‘Integration of wind power in theliberalized Dutch electricity market,’’ Wind Energy, vol. 9, pp. 579–590, 2006.
123
[84] P. Pinson, C. Chevallier, and G. Kariniotakis, ‘‘Trading Wind Generation FromShortTerm Probabilistic Forecasts of Wind Power,’’ IEEE Transactions on
Power Systems, vol. 22, no. 3, pp. 1148–1156, 2007.
[85] J. M. Angarita and J. G. Usaola, ‘‘Combining hydrogeneration and windenergy Biddings and operation on electricity spot markets,’’ Electric Power
Systems Research, vol. 77, pp. 393–400, 2007.
[86] K. Hedman and G. Sheble, ‘‘Comparing Hedging Methods for Wind Power:Using Pumped Storage Hydro Units vs. Options Purchasing,’’ in International
Conference on Probabilistic Methods Applied to Power Systems, 2006., 2006.
[87] J. Matevosyan and L. Soder, ‘‘Minimization of imbalance cost trading windpower on the shortterm power market,’’ IEEE Transactions on Power Sys
tems, vol. 21, no. 3, pp. 1396–1404, 2006.
[88] J. Morales, A. Conejo, and J. PerezRuiz, ‘‘ShortTerm Trading for a WindPower Producer,’’ IEEE Transactions on Power Systems, vol. 25, no. 1, pp.554–564, 2010, 08858950.
[89] J. Barton and D. Infield, ‘‘Energy storage and its use with intermittent renewable energy,’’ IEEE Transactions on Energy Conversion, vol. 19, no. 2, pp.441 – 448, 2004.
[90] A. Cavallo, ‘‘Controllable and affordable utilityscale electricity from intermittent wind resources and compressed air energy storage (CAES),’’ Energy,vol. 32, no. 2, pp. 120–127, 2007.
[91] EPRI, ‘‘EPRIDOE Handbook Supplement of Energy Storage for GridConnected Wind Generation Applications.’’ Electric Power Research Institute,Palo Alto, CA, Tech. Rep., 2004.
[92] ——, ‘‘Electricity Energy Storage Technology Options ,’’ Electric PowerResearch Institute, Palo Alto, CA, Tech. Rep., 2010. [Online].Available: http://www.electricitystorage.org/images/uploads/docs/EPRI_StorageReport_5_11.pdf
[93] H. Ibrahim, A. Ilinca, and J. Perron, ‘‘Energy storage systems–Characteristicsand comparisons,’’ Renewable and Sustainable Energy Reviews, vol. 12,no. 5, pp. 1221–1250, 2008.
[94] M. Korpaas, A. Holen, and R. Hildrum, ‘‘Operation and sizing of energy storage for wind power plants in a market system,’’ International Journal of Elec
trical Power & Energy Systems, vol. 25, no. 8, pp. 599–606, 2003.
[95] R. Walawalkar, J. Apt, and R. Mancini, ‘‘Economics of electric energy storagefor energy arbitrage and regulation in new york,’’ Energy Policy, vol. 35, no. 4,pp. 2558 – 2568, 2007.
124
[96] J. Eyre and G. Corey, ‘‘Energy Storage for the Electricity Grid: Benefits and Market Potential Assessment Guide,’’Sandia National Laboratory, Tech. Rep., 2010. [Online]. Available: http://www.smartgrid.gov/sites/default/files/pdfs/sandia_energy_storage_report_sand20100815.pdf
[97] R. Sioshansi, P. Denholm, and T. Jenkin, ‘‘A comparative analysis of thevalue of pure and hybrid electricity storage,’’ Energy Economics, vol. 33, pp.56–66, 2011.
[98] D. J. Swider, ‘‘Compressed air energy storage in an electricity system withsignificant wind power generation,’’ IEEE Transactions on Energy Conversion,vol. 22, no. 1, pp. 95 –102, 2007.
[99] R. Loisel, A. Mercier, C. Gatzen, N. Elms, and H. Petric, ‘‘Valuation frameworkfor large scale electricity storage in a case with wind curtailment,’’ Energy
Policy, vol. 38, no. 11, pp. 7323–7337, 2010.
[100] P. Denholm and R. Sioshansi, ‘‘The value of compressed air energy storagewith wind in transmissionconstrained electric power systems,’’ Energy Pol
icy, vol. 37, no. 8, pp. 3149–3158, 2009.
[101] H. Lund, G. Salgi, B. Elmegaard, and A. Andersen, ‘‘Optimal operation strategies of compressed air energy storage (CAES) on electricity spot markets withfluctuating prices,’’ Applied Thermal Engineering, vol. 29, no. 56, pp. 799–806, 2009.
[102] S. Succar and R. H. Williams, ‘‘Compressed air energy storage: Theory,resources, and applications for wind power,’’ Princeton EnvironmentalInstitute, Princeton University, Tech. Rep., 2008. [Online]. Available:http://www.princeton.edu/∼ssuccar/caesReport.html
[103] M. Nakhamkin, R. H. Wolk, S. van der Linden, and M. Patel, ‘‘New compressed air energy storage concept improves the profitability of existing simple cycle, combined cycle, wind energy, and landfill gas power plants,’’ inProc. ASME Turbo Expo 2004: Power for Land, Sea, and Air, vol. 5, no. 41707.ASME, 2004, pp. 103–110.
[104] A. Conejo, F. Nogales, and J. Arroyo, ‘‘Pricetaker bidding strategy underprice uncertainty,’’ IEEE Transactions on Power Systems, vol. 17, no. 4, pp.1081–1088, 2002.
[105] A. David and F. Wen, ‘‘Strategic bidding in competitive electricity markets: aliterature survey,’’ in proceedings of IEEE Power Engineering Society SummerMeeting, 2000, vol. 4, 2000, pp. 2168 –2173.
[106] M. Plazas, A. J. Conejo, and F. Prieto, ‘‘Multimarket optimal bidding for apower producer,’’ Power Systems, IEEE Transactions on, vol. 20, no. 4, pp.2041–2050, 2005.
125
[107] P. Attaviriyanupap, H. Kita, E. Tanaka, and J. Hasegawa, ‘‘New bidding strategy formulation for dayahead energy and reserve markets based on evolutionary programming,’’ International Journal of Electrical Power & Energy
Systems, vol. 27, no. 3, pp. 157–167, 2005.
[108] H. Song, C. C. Liu, J. Lawarree, and R. W. Dahlgren, ‘‘Optimal electricitysupply bidding by Markov decision process,’’ IEEE Transactions on Power
Systems, vol. 15, no. 2, pp. 618–624, 2000.
[109] F. Heredia, M. Rider, and C. Corchero, ‘‘Optimal Bidding Strategies for Thermal and Generic Programming Units in the DayAhead Electricity Market,’’IEEE Transactions on Power Systems, vol. 25, no. 3, pp. 1504–1518, 2010.
[110] J. M. Morales, R. Mınguez, and A. J. Conejo, ‘‘A methodology to generatestatistically dependent wind speed scenarios,’’ Applied Energy, vol. 87, no. 3,pp. 843–855, 2010.
[111] A. Prekopa, Stochastic Programming. Norwell: Kluwer Academic Publishers,1995.
[112] J. Birge and F. Louveaux, Introduction to stochastic programming. New York:Springer Verlag, 1997.
[113] P. Kall and J. Mayer, Stochastic Linear Programming: Models, Theory, and
Computation. New York: Kluwer Academic Publishers, 2005.
[114] J. GarciaGonzalez, E. Parrilla, and A. Mateo, ‘‘Riskaverse profitbased optimal scheduling of a hydrochain in the dayahead electricity market,’’ Euro
pean Journal of Operational Research, vol. 181, no. 3, pp. 1354–1369, 2007,doi: DOI: 10.1016/j.ejor.2005.11.047.
[115] J. Kettunen, A. Salo, and D. Bunn, ‘‘Optimization of Electricity Retailer’sContract Portfolio Subject to Risk Preferences,’’ IEEE Transactions on Power
Systems, vol. 25, no. 1, pp. 117–128, 2010.
[116] S. W. Wallace and S. Fleten, ‘‘Stochastic programming models in energy,’’ inHandbooks in Operations Research and Management Science, A. Ruszczynskiand A. Shapiro, Eds. Elsevier Science, 2003, vol. 10, ch. 10, pp. 637–677.
[117] G. B. Dantzig, ‘‘Linear Programming under Uncertainty,’’ Management Sci
ence, vol. 1, no. 3/4, pp. 197–206, 1955.
[118] R. Dahlgren, C. Liu, and J. Lawarree, ‘‘Risk assessment in energy trading,’’IEEE Transactions on Power Systems, vol. 18, no. 2, pp. 503–511, 2003.
[119] M. Denton, A. Palmer, R. Masiello, and P. Skantze, ‘‘Managing market riskin energy,’’ IEEE Transactions on Power Systems, vol. 18, no. 2, pp. 494–502,2003.
126
[120] R. Rockafellar and S. Uryasev, ‘‘Optimization of conditional valueatrisk,’’Journal of Risk, vol. 2, pp. 21–42, 2000.
[121] P. Artzner, F. Delbaen, J. Eber, and D. Heath, ‘‘Coherent measures of risk,’’Mathematical Finance, vol. 9, no. 3, pp. 203–228, 1999, 14679965.
[122] G. Pflug, ‘‘Some remarks on the valueatrisk and the conditional valueatrisk,’’ in Probabilistic Constrained Optimization: Methodology and Applica
tions, S. Uryasev, Ed. Kluwer Academic Publishers, 2000, ch. 1, pp. 1–11.
[123] R. Rockafellar and S. Uryasev, ‘‘Conditional valueatrisk for general lossdistributions,’’ Journal of Banking & Finance, vol. 26, no. 7, pp. 1443–1471,2002, 03784266 doi: DOI: 10.1016/S03784266(02)002716.
[124] J. Cabero, A. Baillo, S. Cerisola, M. Ventosa, A. GarciaAlcalde, F. Peran,and G. Relano, ‘‘A mediumterm integrated risk management model for ahydrothermal generation company,’’ IEEE Transactions on Power Systems,vol. 20, no. 3, pp. 1379–1388, 2005.
[125] R. Jabr, ‘‘Robust selfscheduling under price uncertainty using conditionalvalueatrisk,’’ IEEE Transactions on Power Systems, vol. 20, no. 4, pp. 1852–1858, 2005.
[126] I. Vehvilainen and J. Keppo, ‘‘Managing electricity market price risk,’’ Euro
pean Journal of Operational Research, vol. 145, no. 1, pp. 136–147, 2003.
[127] AESO, ‘‘Current and historical market reports,’’ 2011. [Online]. Available:http://ets.aeso.ca
[128] ——. Market reports. [Online]. Available:http://www.aeso.ca/downloads/Market_Reports_Jan_2010.pdf
[129] EIA, ‘‘Electricity market module,’’ Energy Information Administration, Tech. Rep., 2010. [Online]. Available:http://www.eia.doe.gov/oiaf/aeo/assumption/pdf/electricity.pdf
[130] ERCB, ‘‘Alberta’s Energy Reserves 2010 and Supply/Demand Outlook20112020,’’ Energy Resources and Conservation Board, Tech. Rep. ST98,2011. [Online]. Available: http://www.ercb.ca/docs/products/STs/st98_current.pdf
[131] N. LeBlanc, D. McColl, G. Eynon, A. Naini, and M. Stogran, ‘‘Cogenerationopportunities and energy requirements for Canadian oil sands projects part III: Cogeneration options and opprotunities,’’ Canadian Energy ResearchInstitute, Tech. Rep., 2005.
[132] ACR, ‘‘Oil sands technology roadmap: Unlocking the potential,’’Alberta Chamber of Resources, Tech. Rep., 2004. [Online]. Available:http://www.acralberta.com/OSTR_report.pdf
127
[133] J. Moorhouse and B. Peachey, ‘‘Cogeneration and the Alberta oil sands,’’Cogeneration & Onsite Power Production, vol. 8, no. 4, pp. 121–127, 2007.
[134] ERCB, ‘‘Alberta crude bitumen in situ production monthly statistics,’’ EnergyResources and Conservation Board, Tech. Rep. ST53, 2009.
[135] OSDG, ‘‘Oil sands cogeneration report 2010,’’ Oil Sands Developers Group, Survey Report, 2010. [Online]. Available:http://www.oilsandsdevelopers.ca/index.php/library/
[136] AESO, ‘‘Long term adequacy metrics,’’ 2010. [Online]. Available:http://www.aeso.ca/market/19869.html
[137] AENV, ‘‘Alberta’s 2008 climate change strategy,’’ Alberta Environment, Tech. Rep., 2008. [Online]. Available:http://environment.gov.ab.ca/info/library/7894.pdf
[138] ——, ‘‘Specified gas emitters regulation: technical guidance document forbaseline emissions intensity applications,’’ Alberta Environment, Tech. Rep.,2009. [Online]. Available: http://environment.alberta.ca/01094.html
[139] ——, ‘‘Specified gas emitters regulation: additional guidance on cogenerationfacilities,’’ Alberta Environment, Tech. Rep., 2007. [Online]. Available:http://environment.gov.ab.ca/info/library/7930.pdf
[140] J. A. Jacobs and M. Schneider, ‘‘Cogeneration application considerations,’’ GE Energy, Tech. Rep., 2009. [Online]. Available: http://www.gepower.com/prod_serv/products/tech_docs/en/downloads/GER3430G.pdf
[141] N. LeBlanc, D. McColl, A. Naini, L. Chan, M. Stogran, C. Bramley, andB. Dunbar, ‘‘Cogeneration opportunities and energy requirements for Canadian oil sands projects part I: Status of oil sands developments 2005,’’Canadian Energy Research Institute, Tech. Rep., 2005.
[142] T. Ekvall, ‘‘Allocation in ISO 14041—a critical review,’’ Journal of Cleaner
Production, vol. 9, no. 3, pp. 197–208, 2001.
[143] J. B. Guinee, Ed., Handbook on Life Cycle Assessment: Operational Guide to
the ISO Standards. Dordrecht: Kluwer Academic Publishers, 2002.
[144] D. T. Allen, C. Allport, K. Atkins, J. S. Cooper, R. M. Dilmore, L. C.Draucker, K. E. Eickmann, J. C. Gillen, W. Gillette, W. M. Griffin, W. E.Harrison, J. I. Hileman, J. R. Ingham, F. A. Kimler, A. Levy, C. F. Murphy,M. J. O’Donnell, D. Pamplin, G. Schivley, T. J. Skone, S. M. Strank, R. W.Stratton, P. H. Taylor, V. M. Thomas, M. Q. Wang, and T. Zidow, ‘‘Frameworkand guidance for estimating greenhouse gas footprints of aviation fuels,’’Airforce Research Laboratory, Interim Report, 2009. [Online]. Available:http://www.netl.doe.gov/energyanalyses/pubs/EstGHGFtprntsAvFuels2009.pdf
128
[145] J. B. Guinee and R. Heijungs, ‘‘Calculating the influence of alternative allocation scenarios in fossil fuel chains,’’ The International Journal of Life Cycle
Assessment, vol. 12, no. 3, pp. 173–180, 2006.
[146] M. A. Curran, ‘‘Studying the Effect on System Preference by Varying Coproduct Allocation in Creating LifeCycle Inventory,’’ Environmental Science
& Technology, vol. 41, no. 20, pp. 7145–7151, 2007.
[147] E. Gnansounou, A. Dauriat, and J. Villegas, ‘‘Bioresource Technology : Lifecycle assessment of biofuels: Energy and greenhouse gas balances,’’ Biore
source technology, 2009.
[148] S. Suh, B. Weidema, J. H. Schmidt, and R. Heijungs, ‘‘Generalized Make andUse Framework for Allocation in Life Cycle Assessment,’’ Journal of Industrial
Ecology, vol. 14, no. 2, pp. 335–353, 2010.
[149] J. B. Guinee, R. Heijungs, and E. Voet, ‘‘A greenhouse gas indicator for bioenergy: some theoretical issues with practical implications,’’ The International
Journal of Life Cycle Assessment, vol. 14, no. 4, pp. 328–339, 2009.
[150] M. A. Rosen, ‘‘Allocating carbon dioxide emissions from cogeneration systems: descriptions of selected outputbased methods,’’ Journal of Cleaner
Production, vol. 16, no. 2, pp. 171–177, 2008.
[151] R. Frischknecht, ‘‘Allocation in life cycle inventory analysis for joint production,’’ The International Journal of Life Cycle Assessment, vol. 5, no. 2, pp.85–95, 2000.
[152] B. P. Weidema and J. H. Schmidt, ‘‘Avoiding Allocation in Life Cycle Assessment Revisited,’’ Journal of Industrial Ecology, vol. 14, no. 2, pp. 192–195,2010.
[153] AESO, ‘‘2008 Annual report,’’ Alberta Electric System Operator, Calgary, 2009. [Online]. Available: http://www.aeso.ca/downloads/20100415_AESO_2009_DA_Reconciliation__Appendix_D2__2008_Annual_Report.pdf
[154] MSA, ‘‘Weekly market monitor reports,’’ 2009. [Online]. Available:http://www.albertamsa.ca
[155] AENV, ‘‘Greenhouse Gas Reporting Program,’’ 2011. [Online]. Available:http://environment.alberta.ca/02166.html
[156] AESO. (2010) Long term adequacy metrics. [Online]. Available:http://www.aeso.ca/market/21311.html
129
Appendix A
List of Power Generating Units
Generating units in the power system simulation models developed in chapter 2are listed in table A.1.
Table A.1: Power generating units
Unit ID Bus Pmin Pmax Type a Heat rateb Ramp rate Start up fuel Min. up time Min. down time(MW) (MW) (GJ/MWh) (MW/min) (GJ) (hours) (hours)
101 3 60 148 PC 12 2 2500 48 24102 3 60 148 PC 12 2 2500 48 24103 3 147 368 PC 11 3 3000 48 24104 4 154 384 PC 10 3 3000 48 24105 4 154 384 PC 10 3 3000 48 24106 4 158 450 SCPC 9 5 3000 48 24107 6 60 143 PC 16 2 3000 48 24108 4 152 381 PC 10 3 3000 48 24109 4 152 381 PC 10 3 3000 48 24110 1 151 378 PC 11 3 3000 48 24111 1 151 378 PC 11 3 3000 48 24112 4 112 280 PC 11 3 3000 48 24113 4 112 280 PC 11 3 3000 48 24114 4 141 353 PC 11 3 3000 48 24115 4 162 406 PC 11 3 3000 48 24116 4 141 353 PC 10 3 3000 48 24117 4 160 399 PC 10 3 3000 48 24118 4 112 279 PC 12 2 3000 48 24210 2 10 250 CCGT 8 12 375 4 2215 6 5 25 SCGT 13 5 65 1 1207 4 9 46 SCGT 13 5 120 1 1209 1 5 27 SCGT 13 6 70 1 1226 6 10 48 SCGT 13 5 125 1 1228 6 5 26 SCGT 17 3 88 1 1229 6 8 40 SCGT 16 4 128 1 1230 6 4 21 SCGT 16 3 67 1 1241 6 9 45 SCGT 12 5 108 1 1201 5 37 184 Cogen 7.5 7 49 4 3202 5 16 80 Cogen 7.5 3 21 4 3203 6 10 50 Cogen 7.5 2 13 4 3204 6 6 30 Cogen 7.5 1 8 4 3205 3 1 6 Cogen 7.5 1 2 4 3206 3 19 95 Cogen 7.5 4 25 4 3208 5 65 326 Cogen 7.5 13 86 4 3211 6 2 12 Cogen 7.5 1 3 4 3212 2 24 120 Cogen 7.5 5 32 4 3213 6 9 47 Cogen 7.5 2 12 4 3214 5 16 80 Cogen 7.5 3 21 4 3216 3 95 474 Cogen 7.5 15 125 4 3217 5 33 165 Cogen 7.5 7 44 4 3218 5 36 180 Cogen 7.5 7 47 4 3222 1 48 239 Cogen 7.5 10 63 4 3223 5 40 200 Cogen 7.5 8 53 4 3224 2 24 120 Cogen 7.5 5 32 4 3225 5 36 180 Cogen 7.5 7 47 4 3227 5 17 85 Cogen 7.5 3 22 4 3231 6 9 47 Cogen 7.5 2 12 4 3232 6 9 47 Cogen 7.5 2 12 4 3233 5 8 40 Cogen 7.5 2 11 4 3234 4 4 19 Cogen 7.5 1 5 4 3
Continued on next page
130
Table A.1 – continued from previous page
Unit ID Bus Pmin Pmax Type a Heat rateb Ramp rate Start up fuel Min. up time Min. down time(MW) (MW) (GJ/MWh) (MW/min) (GJ) (hours) (hours)
237 5 105 525 Cogen 7.5 15 138 4 3238 5 102 510 Cogen 7.5 19 135 4 3239 4 2 11 Cogen 7.5 1 3 4 3240 4 8 39 Cogen 7.5 2 10 4 3301 3 0 120 Hydro 24 0 0 0302 2 0 320 Hydro 64 0 0 0303 3 0 350 Hydro 70 0 0 0304 1 0 89 Hydro 6 0 0 0501 6 10 99 Biomass 12 1.7 0 4 3502 1 7 36 Biomass 12 0.6 0 4 3503 3 2 11 Biomass 12 0.2 0 4 3504 6 5 27 Biomass 12 0.5 0 4 3507 6 4 18 Biomass 12 0.3 0 4 3508 6 5 25 Biomass 12 0.4 10 4 3401 1 0 500 Wind 0 0 0
a PC Pulverized coal; SCPC Super critical pulverized coal; CCGT Combined cycle gas turbine; SCGT Simple cycle gasturbine; Cogen Cogeneration (natural gas fired)b Higher heating value (HHV) basis
131
Appendix B
Optimal Operation of Standalone Wind PowerGeneration System
Nomenclature
Sets, indices, and parameters
S, s Set and index of scenariosT, t Set and index of timeω1, ω2 Weighting factors that set the risk preferenceτ Optimization time step (=1h)ρs Probability of scenario sπst Market price under scenario s in hour t [$/MWh]Wst Output of the wind power system under scenario s in hour t [MW]λ Penalty factor for energy imbalancePwmax Maximum power generation limit of the wind farm [MW]Ptx Maximum available transmission capacity [MW]Pramp Maximum allowable ramp rate [MW/h]
Decision variables
gwst Power output of the wind farm in hour t under scenario s [MW]bwt Energy bid to the day ahead market by the wind farm in hour t [MWh]αw Profit threshold level [$]
Model Formulation
The objective of the operator of the standalone wind power generation system,represented by the objective function (B.1), is to maximize the profits earned byselling wind energy, taking into account any imbalance settlements under eachscenario s. Wind energy sold, gwsh, in each scenario is constrained by the windenergy production level in the respective scenario, Wsh (B.2). Hourly energy bids,bwt to the dayahead market are constrained by the installed wind capacity (B.3).Constraint (B.4) represents the transmission limitation. The interhour rampinglimit is enforced by (B.5). The model can be implemented as a linear programmingproblem (LP).
132
maximize ω1(∑
s∈S
ρsBws ) + ω2(α
w −1
(1− β)
∑
s∈S
ρs · [αw − Bw
s ]+) (B.1)
where, Bws = (πstg
wst · τ − λ · πst|g
wst · τ − bwt |); ω1, ω2 ≥ 0
subject to :
0 ≤ gwst ≤ Wsh, ∀s ∈ S, ∀t ∈ T (B.2)
0 ≤ bwt ≤ Pwmax · τ, ∀s ∈ S, ∀t ∈ T (B.3)
0 ≤ gwst ≤ Ptx, ∀s ∈ S, ∀t ∈ T (B.4)
0 ≤ |gwst − gwst−1| ≤ Pramp, ∀s ∈ S, ∀t ∈ T (B.5)
133
Appendix C
SGER Obligations Calculations
The total fuel consumed, electricity generated/imported, and CO2 emissions underthe ‘‘Baseline Option" and the ‘‘Cogeneration Option" are calculated using equations(4.1)(4.5) listed in section 4.3. These results are listed in tables C.1 and C.2.
Table C.1: Mass and energy balances of the ‘‘Baseline option"
Hourly amount Annual amount
Electricity Imports, E 15 MWh 120 GWhSteam enthalpy, H 1600 GJ 124400 TJBoiler feed water enthalpy, Hfw 480 GJ 4 TJBoiler fuel input, FB 1300 GJ 10400 TJBitumen production, PB 1250 bbl 1000000 bbl
Table C.2: Mass and energy balances of the ‘‘Cogeneration option"
Hourly amount Annual amount
Electricity production, Ec 83 MWh 650 GWhElectricity exports, Eexp 68 MWh 530 GWhGas turbine fuel input, FT 1020 GJ 8000TJHRSG steam enthalpy, H1 1100 GJ 8800 TJHRSG feed water enthalpy, Hfw1 330 GJ 2600 TJHRSG fuel input, FG 450 GJ 3700 TJSupplementary boiler steam enthalpy, H2 590 GJ 3800 TJSupplementary boiler feed water enthalpy, Hfw2 150 GJ 1100 TJSupplementary boiler fuel input, FSB 400 GJ 3100 TJBitumen production, PC 1250 bbl 1000000 bbl
Emissions reduction obligation under Alberta’s Specified Gas Emitters Regulation (SGER) is calculated using the guidelines set by Alberta Environment[138,139]. We assume that the facility represented in the model commenced itscommercial operations after 2000 (therefore, considered as a ‘‘new facility" underSGER) and has completed more than 8 years of operations. We also assume thatthe facility’s bitumen production and fuel consumption does not significantly varyfrom the baseline year (According to SGER guidelines the baseline year for a ‘‘newfacility" is the third year of its commercial operations). For simplicity we calculateonly the emissions reduction obligations associated with satisfying the facilities
134
steam and electricity demand. Furthermore, we consider only the CO2 emissions,excluding other GHGs. The SGER emissions reduction obligation (ERO) is calculated using equation (C.1) [139].
ERO = TAE − BEI · (1− t) · P (C.1)
where, TAE = Total annual GHG emissions (Jan 1Dec 31)
BEI = Baseline emissions intensity
t = Emissions reduction target (currently 12%)
P = Facility’s annual bitumen production (Jan 1Dec 31)
If ERO is positive, the facility must submit compliance options. If ERO is negative the facility has earned ‘‘Emissions Performance Credits (EPC)". EPCs may besold to parties with reduction obligations or be banked or future use.
The BEI and TAE of the ‘‘baseline option" are calculated using equations (C.2)and (C.3) respectively.
BEIabseline =TAEbaseline year
Pbaseline year
(C.2)
TAE = FB · Icng (C.3)
where, Icng = CO2 emissions intensity of natural gas (assumed to be 0.0503 tCO2/GJ)
SGER recognizes the energy efficiency gains achievable through cogenerationand attempt to incentivize a facility with cogeneration by requiring to comply onlyfor the emissions associated with thermal energy production. The BEI and EROcalculations are adjusted to exclude the emissions associated with electricity production. In this analysis we use the SGER guidelines set for an ‘‘Integrated Cogeneration Facility". Integrated cogeneration facilities are those that, in addition tothe cogeneration units, also have other means of producing thermal energy and/orelectricity (in our illustrative example system, under the cogeneration option thereis a supplementary boiler to produce additional steam). Under SGER the following
assumptions are made in calculating BAI and TAE of a cogeneration facility:
1. In the baseline year, thermal energy is generated by a boiler with a thermalefficiency of 80% (SGER guidelines does not specify whether this is in higherheating value (HHV) or in lower heating value (LHV); we assume HHV for ourcalculations).
2. In the absence of the cogeneration unit, the power producer would have tobuild a natural gas fired combined cycle gas turbine (CCGT) based powerplant to satisfy the facilities electricity demand. Therefore the GHG emissionsintensity of the cogenerated electricity is equal to 0.418tCO2/MWh, which isthe GHG intensity of the CCGT power plant.
135
The BEI of the integrated cogeneration facility is calculated using the equations(C.7)(C.7):
Gt = (Ft + FG) · Icng (C.4)
TAEcogen = (FT + FG + FSB) · Icng (C.5)
DH =H1 −Hfw1
0.8· Icng (C.6)
BEIcogen =(TAEcogen(baseline year) −Gt(baseline year)) +DH
Pbaseline year
(C.7)
where, Gt = Annual CO2 emissions from the fuel consumed by the cogeneration unit
DH = Deemed CO2 emissions from heat production
In ERO calculations, the emissions associated with electricity are excluded bysubtracting out ‘‘deemed CO2 emissions from electricity generation, DE". Equations(C.8)(C.9) are used to calculate ERO.
DE = 0.418EC (C.8)
EROcogen = (TAEcogen −DE)− BEIcogen · (1− t) · P (C.9)
Results of SGER obligations calculations are listed in table C.3. In this example, under the ‘‘cogeneration option" the facility has earned EPCs that amounts to15000 tCO2.
Table C.3: SGER obligations
Baseline option Cogeneration option
Baseline emissions intensity (BEI) 0.05 tCO2/bbl 0.06 tCO2/bblTotal annual emissions (TAE) 520 ktCO2 740 ktCO2
Emissions reduction obligation (ERO) 63 ktCO2 15 ktCO2
136
Appendix D
Alberta Grid Average and Marginal Emissions IntensityCalculations
Average and marginal CO2 emissions intensity of the Alberta Electric System is calculated using the data published by the Alberta Electric Systems Operator (AESO).The amount of generation in Alberta by generation technology for the period 20002008 is listed in Table D.1 [153]. CO2 intensity of each generation technology islisted in Table D.2.
Table D.1: Alberta’s electricity production by generation technology (in GWh)
Coal Gas Cogeneration Hydro Wind Other Imports
2000 40,885 8,477 2,699 1,687 36 275 1,3072001 41,753 5,011 5,662 1,424 112 261 9082002 42,853 2,382 6,736 1,650 209 311 1,1352003 41,608 1,781 8,433 1,717 341 315 1,3282004 42,203 2,071 8,369 1,930 549 316 1,4922005 43,905 1,760 7,298 2,324 764 373 1,5352006 44,576 1,928 7,917 1,804 860 490 1,5172007 44,191 1,734 8,509 2,050 1,451 497 1,4672008 42,270 1,676 8,148 2,025 1,543 358 2,248
Source: [153]
Table D.2: CO2 intensity of the generation technology (in tCO2/MWh)
Year Coal Gas Cogeneration Hydro Wind Other Imports
CO2 intensity 1.034 0.5 0.19 0.41 0 0 0 0
Table D.2 notes:
i CO2 emissions intensity of coal fired electricity is calculated by assuming the average heat rate of coal fired generation units in Alberta to be 11.4GJ/MWh [48] and the CO2 intensity of coal to be 0.09 tCO2/GJ [56]
ii The ‘‘Gas" category includes both CCGT and SCGT based natural gas fired generation and a weighted average CO2 intensity is used for the calculations.Heat rate data of the natural gas fired generation units is obtained from [48]. The CO2 intensity of natural gas is assumed to be 0.05 tCO2/GJ [56]
iii Cogeneration emissions intensities are calculated as described in the section 4.3.
iv The ‘‘Other" category is mainly consists of biomass based generation and therefore zero CO2 intensity is assumed
v Zero carbon intensity is assumed for imports as the generation takes place outside of Alberta
The average CO2 intensity, Ic_ABav of a given year is calculated using equation(D.1) and the data in tables D.1 and D.2. Figure D.1 depicts the calculated Ic_ABav
values for the period 20002008.
137
Ic_ABav =
∑
Ef · Icf∑
Ef
(D.1)
where, f = coal, gas, cogeenration, hydro, wind, other, imports
Ef = Amount of electricity generated by technology f in a given year
Icf = CO2 intensity of the technology f
As discussed in sections 4.4.1 and 4.4.2, CO2 intensity of cogenerated electricitydepends on the allocation method and therefore the grid average CO2 (and themarginal emissions intensity) too depends on the allocation method. Variationsin the gird average CO2 intensity due to the allocation method used is depicted inFigure S1.
2000 2001 2002 2003 2004 2005 2006 2007 20080
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Ave
rage
CO
2 inte
nsity
, tC
O2/M
Wh
Year
Figure D.1: Average CO2 intensity of the Alberta Grid in years 2000 to 2008Light blue areas depict the variations of the grid average CO2 intensity of each yeardue to the allocation method used to divide fuel between cogenerated steam andelectricity.
We calculate the marginal CO2 emissions intensity of the Alberta grid, Ic_ABom
assuming that the marginal unit of the generation stack is the price setting generator. In order to calculate the marginal CO2 intensity accurately, detailed generationdispatch information such as the time each generator operates at the margin andtheir heat rates are required. However, this data is kept confidential in a competitive power market environment as in the case of Alberta. Therefore we use anaggregated data set published by the AESO [153] to calculate the marginal CO2
intensity of the Alberta grid. This data set is listed in Table D.3 and specifies thepercentage of time each generation technology sets the system price from 2000to 2008. The marginal CO2 intensity, calculated by equation (D.2), is the timeweighted sum of the CO2 intensities of generation technologies that are listed intable D.2.
138
Table D.3: Percentage of the time different generation technologies set the price inAlberta’s whole sale electricity market.
Year Coal Cogeneration Gas Hydro Import Load
2000 12% 5% 45% 9% 29% 1%2001 35% 14% 47% 4% 0% 0%2002 47% 22% 29% 3% 0% 0%2003 49% 26% 24% 1% 0% 0%2004 46% 17% 36% 1% 0% 0%2005 57% 25% 16% 2% 0% 0%2006 59% 23% 17% 1% 0% 0%2007 68% 22% 10% 1% 0% 0%2008 50% 34% 15% 1% 0% 0%
Ic_ABom =∑
Mf · Icf (D.2)
where, f = coal, gas, cogeenration, hydro, wind, other, imports
Mf = Percentage of the time the generation technology f sets the system price
Icf = CO2 intensity of the technology f
The marginal CO2 intensities of the Alberta grid for the period 20002008 areshown in figure D.2. As can be seen from that figure and table D.3 the amountof time cogeneration sets the system price is increasing and therefore the effect ofthe allocation method on the marginal CO2 intensity too has increased during thisperiod.
2000 2001 2002 2003 2004 2005 2006 2007 20080
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Mar
gina
l CO
2 inte
nsity
, tC
O2/M
Wh
Year
Figure D.2: Marginal CO2 intensity of the Alberta gridLight blue areas depict the variations of the grid average CO2 intensity of each yeardue to the allocation method used to divide fuel between cogenerated steam andelectricity.
139
Appendix E
CO2 Emissions Forecast of the Alberta Electric System
In this section we present the modelling details of the Alberta electricity sectorCO2 emissions forecast to 2020 presented in section 4.4.1 (Figures 4.4 & 4.5).This forecast was developed considering the present fleet of generators, plannedgeneration additions and retirements, and effective generation capacity required tomeet the forecasted demand growth in Alberta to the year 2020. According to theAESO’s forecast, peak demand in Alberta in 2020 will be 15350 MW. In order tomaintain system reliability the AESO requires a 10% generation reserve margin.Therefore by 2020 the Alberta Electric System requires an installed generationcapacity of 16885 MW. Because of the intermittency of wind and hydro generatorstheir capacity should be derated to determine the effective generation capacityavailable to satisfy the peak load. The AESO derates the installed wind capacity to20% and hydro capacity to 5067% [46]. In this forecast we derate wind capacity to20% and hydro capacity to 62% (derating factor was chosen based on the installedcapacities of reservoir based hydro units versus runoftheriver hydro units). TableE.1 lists the installed generation capacity at the end of 2009, the effective generationcapacity after derating wind and hydro, and planned generation retirements andadditions between 2010 and 2020. The majority of the planned unit retirementsare older coal units due to the expiration of power purchase agreements (PPA) inthe period of 20172020. Currently these coal units mainly satisfy baseload. Inour model the generation projects that are under active construction and thosewith regulatory approval as planned additions are included. After adjusting forthe retirements, additions and derating wind and hydro installed capacity, theeffective generation capacity in Alberta in 2020 is 12059MW, 4826MW short ofsatisfying the forecasted peak demand in 2020.
We consider three generation scenarios to meet the capacity shortage mentionedabove. Factors that are taken into account to develop these scenarios include theresource availability, generation projects that have applied for regulatory approval[136], forecasted growth of the oil sands sector, and future electricity generationscenarios used by the AESO in its long term transmission expansion plan [46].The three scenarios are listed in table E.2. The assumptions made to formulate thescenarios are described below:
• Considering the generation projects that have applied for regulatory approvaland the AESO’s generation forecast, all three scenarios include 350MW ofCCGT, 600MW of SCGT and 1200MW of wind.
• Every scenario includes 550MW of new coal fired generation. This includes a450MW brownfield unit addition and expansion of the capacity of the existingunits totalling 100MW.
• All three scenarios include 1300MW of new cogeneration additions. This
140
includes the capacity expansion of existing cogeneration systems; integrationof cogeneration into existing oil sands operations based on oil sands projectoperators announcements.
• Scenario GS1 includes an additional four 450MW coal fired generation unitsmaking the total coal capacity in GS1 2350MW. This scenario is plausibledue to the large amount of coal reserves in the northcentral region of theprovince closer to existing coal fired generation sites and the new transmission expansions announced by the AESO.
• Scenario GS2 includes an additional 1800MW of new cogeneration. This isalso plausible due to the forecast growth in oil sands projects as described insection 2.1. The total cogeneration capacity in GS2 is 2100MW.
• Scenario GS3 includes an additional 1800MW of new CCGT capacity makingthe total CCGT capacity 2150MW. Under strict carbon control regulation, itis plausible that the baseload generation will be dominated by CCGT units.Two proposed CCGT projects with a total capacity of 1150MW have receivedregulatory approval [136].
Table E.1: Installed electricity generation capacity in Alberta (in MW)
Coal CCGT SCGT Cogen Hydro WindBiomass& other Total
Installed capacity at theend of 2009 5946 337 627 3869 871 563 284 12497
Planned Retirements(20102020) 1096 0 105 0 0 0 0 1201
Planned additions(20102020) 496 0 417 455 100 406 33 1907
Net installed capacity in2020 after retirementsand additions 5346 337 939 4324 971 969 317 13203
Effective generation capacity in 2020 after derating wind and hydro 5346 337 939 4324 602 194 317 12059
Sources: [46,156]
Capacities of these three scenarios are linearly added to the installed capacities listed in table E.1 (Figure E.1). We then assume these generation units willoperate with their typical average utilization factors (ie. coal: 80%, CCGT: 50%,SCGT: 10%, hydro: 40%, wind: 30%, biomass & other: 50%) to calculate annualelectricity generation by each technology. However, in the case of GS3, we assumea capacity factor of 90% for CCGT units as they provide the baseload power. The
141
Table E.2: Generation Scenarios
GenerationScenario1 (GS1)(MW)
GenerationScenario2 (GS2)(MW)
GenerationScenario3 (GS3)(MW)
Coal 2350 550 550
CCGT 350 350 2150
SCGT 600 600 600
Cogeneration 1300 2100 1300
Wind 1200 1200 1200
Total generation capacity 5800 5800 5800
Total effective generation capacity 4840 4840 4840
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 20200
2000
4000
6000
8000
10000
12000
14000
MW
CoalCCGTSCGTCogenerationHydroWindBiomass
Figure E.1: Forecasted installed generation capacity in Alberta (20092020)
annual electricity production of each technology is then multiplied by the emissions intensities listed in Table S2 and summed to estimate the total annual CO2
emissions from electricity generation. The estimated CO2 emissions from electricity generation in the period 20092020 is appended to the actual CO2 emissionsfrom electricity generation in the period 2000 to 2008. Figure 4.4 depicts the CO2
emissions under the generation scenarios GS1 and GS2. The same results underthe generation scenarios GS2 and GS3 are depicted in figure 4.5. As can be seenfrom figure 4.5, the CO2 emissions under the high cogeneration scenario GS2 isonly 25% (depending on the allocation method employed for cogeneration units)lower than the emissions under the high CCGT scenario GS3.